Office Hours

[agcohn821]

9/25/20- led by beta_helix

[9:00 AM] beta_helix: Hi everyone! Welcome to Foldit Office Hours! I'm beta_helix: (http://www.cis.umassd.edu/~fkhatib/cv.html) I've been on the Foldit Team since 2008.
[9:00 AM] Dudit: Hi @beta_helix
[9:00 AM] beta_helix: Hi Dudit!
[9:00 AM] joshmiller: hi beta!
[9:01 AM] beta_helix: I'd love to open the floor to any players that are joining our Foldit Office Hours for the first time
[9:01 AM] beta_helix: Hi Josh!
[9:02 AM] Todd6485577: morning all
[9:02 AM] beta_helix: Good morning/afternoon/evening (depending on where everyone is.
[9:03 AM] jmbrownlee333: For josh, how is the thesis proposal coming. If i dare ask.
[9:03 AM] beta_helix: I want to know that answer too!
[9:04 AM] joshmiller: oh hey jmb! sorry, I'm only half paying attention because I'm in a meeting haha (Seth is there too, we're talking about another cit sci game related to iNaturalist)
[9:05 AM] jmbrownlee333: nice multi-tasking
[9:05 AM] joshmiller: thesis proposal has been submitted, still waiting to hear back from my college on whether I'm allowed to give the proposal. Almost done with data analysis on a Foldit-related paper
[9:05 AM] joshmiller: some interesting results came out of it, lots I expected and hope to fix in future design work
[9:05 AM] beta_helix: Nice! Say hi to Seth, unless you would get in trouble for not paying attention to the meeting
[9:06 AM] joshmiller: :P nah I'll do it
[9:06 AM] beta_helix: sweet
[9:06 AM] beta_helix: I feel like Dudit is typing a novel…
[9:07 AM] Dudit: I think Foldit should have a new achievement in the Foldit Achievement page, called 'Successfull Solution' when any Foldit player solution (regardless the score) created a solution that Really Works in the Wet Lab
[9:07 AM] beta_helix: That is a great idea, Dudit!
[9:07 AM] beta_helix: and perfect timing as Josh has started looking into revamping achievements
[9:08 AM] beta_helix: (I hope I wasn't supposed to keep that a secret, Josh!)
[9:09 AM] jmbrownlee333: will we also see re-vamped categaories or updated category rankings
[9:09 AM] beta_helix: that too, jmb!
[9:09 AM] beta_helix: That should coincide with the new website we hope to launch soon
[9:10 AM] beta_helix: There are just so many things that we can't support/enhance with the current Foldit site… so it requires an entirely new website for many of these features.
[9:10 AM] beta_helix: For example: being able to view another player's solution on the site, and rotate it
[9:11 AM] Josh: haha not a secret beta_helix:, I would love to add that achievement!
[9:12 AM] beta_helix: I feel like that one is even more important than The Write Stuff
[9:12 AM] Dudit: I think there is a problem that Josh had mention it before, called 'Over Approximation', it is when the Foldit player solution scores highest but failed to work in the wet lab
[9:12 AM] beta_helix: Yes indeed, Dudit.
[9:13 AM] beta_helix: Design is very tricky, because the score function isn't as simple as with Electron Density (for example).
[9:14 AM] beta_helix: In ED puzzles, the top scores are usually spot on (or very close) to the native… unless nobody was able to reach the native fold, of course.
[9:14 AM] beta_helix: That is not the case with design, which is why the Share with Scientists feature is so critical!
[9:15 AM] jeff101: I have some questions about hydrogen bonds and disulfide bonds.
[9:15 AM] jeff101: I noticed more Shares wth Scientists on Puzzle 1890.
[9:15 AM] beta_helix: We can't test the hundreds of thousands of Foldit designs that come back, but we can easily check the Scientist Shares
[9:16 AM] jeff101: For disulfide bonds that Foldit shows and gives +250 bonuses for, I've seen negative Disulfide subscores.
[9:16 AM] beta_helix: Hi jeff101! Do you think those are more bug shares or science shares?
[9:16 AM] jeff101: I think Science Shares. Maybe the max is 10 now.
[9:17 AM] beta_helix: Hmmm… have you shared such a case so we can look into it?
[9:17 AM] jeff101:but for hydrogen bonds that Foldit shows as blue and white candy canes, I think the Bonding subscores are always positive.
[9:18 AM] beta_helix: I know that sometimes you can get the disulfide bonus even when the angle isn't great or there is even a clash.
[9:20 AM] jeff101:on 1890, I shared w/myself at some point: Je11-37 8850.298 +750 lo 9/18/20 1242am midnight
[9:20 AM] beta_helix: Ok thanks, I'll let the team know to investigate what is going on with that one.
[9:20 AM] jeff101:it has 3 s-s bonds: 2-7 w/score 6.9, 8-39 w/score -4.4, and 11-37 w/score -3.6
[9:21 AM] spvincent: Are there any tools in the works to better visualize bonding in H bond networks?
[9:21 AM] jeff101:another one, Je0 8887.278 +750 lo 9/22/20 1146pm
[9:21 AM] Dudit: I think Foldit needs more performance optimization
[9:21 AM] jeff101:had 3 s-s bonds: 2-8 w/score 2.3, 7-20 w/score -1.7, and 32-39 w/score 0.
[9:22 AM] beta_helix: @spvincent is there a specific visualization that you'd want?
[9:22 AM] jeff101:There are other cases I could find if you need them.
[9:23 AM] beta_helix: @Dudit yes, that it indeed the case. Especially if we want to be able to post larger proteins without grinding the game down to a halt.
[9:23 AM] spvincent: Better visualization of the unsatisfied bonds
[9:23 AM] beta_helix: @jeff101 Thanks! Hopefully these 2 will help us uncover what is going on.
[9:26 AM] Dudit: I wish the Office Hour is scheduled every week, I believe there are a lot of question that needs to be answered
[9:26 AM] Josh: Me too…
[9:27 AM] jeff101:For h-bonding to a particular atom, is there a limit on how many h-bonds can form? Like a on a bent O acceptor, is the most h-bonds 2 ? Do they sort of form a tetrahedron around the O atom?
[9:27 AM] beta_helix: @spvincent We are working on ways to make it easier to form proper h-bonds… particularly using rubber bands
[9:27 AM] spvincent: ok, nice
[9:27 AM] beta_helix: @Dudit Good news! You can bring this up in our new Foldit Office Hours feedback section: https://fold.it/portal/node/2010431
[9:27 AM] jeff101:Also, for a carbonyl (C=O) oxygen atom, is the most h-bonds to the O acceptor 2? Do they come in 120 degrees apart?
[9:30 AM] agcohn821: Thank you @beta_helix –glad you found that! Yes, please post feedback!
[9:31 AM] beta_helix: @jeff101 This is probably why I should have taken Organic Chemistry… too bad I was a math major in college! I don't want to give you the wrong information, so I will ask our resident biochemists on the team and get back to you on that one. (Unless any of them are online right now and want to save me.
[9:32 AM] beta_helix: @Dudit do you have any other questions that I can help answer today? (just not about O-chem.
[9:33 AM] spvincent: The puzzles have become a bit repetitive of late.
[9:33 AM] jeff101:A harder question is what is the max # of h-bonds to a nitrogen that is an acceptor atom.
[9:33 AM] jeff101:Can nitrogens even be acceptor atoms?
[9:34 AM] beta_helix: @spvincent This is important feedback that we need to hear. Thank you! What kind of puzzles are you missing? Hand-folding?
[9:34 AM] jeff101:perhaps N's on rings like C=N-C can be acceptors
[9:35 AM] Dudit: @beta_helix: Is it true that Foldit only using CPU (not iGPU) because it is using Rosetta?
[9:35 AM] jeff101:would the hydrogen bond be in a plane with the C=N-C atoms, so that all 3 angles are 120 degrees?
[9:36 AM] beta_helix: @jeff101 Thank you! I will make sure to pass all those questions to the biochem experts.
[9:36 AM] beta_helix: @Dudit Yes.
[9:37 AM] beta_helix: I was in the Baker Lab from 2008-2013 and there were a bunch of people that really wanted to push Rosetta to work on GPUs… they put in a lot of time and effort in this.
[9:38 AM] beta_helix: Ultimately, the Rosetta Community did not go in that direction.
[9:39 AM] spvincent Not hand-folding specifically. But maybe some more small molecule binder stuff: Symmetric puzzles with different numbers of monomers (what is so special about the number 3): revisiting aflatoxin: etc. Ed puzzles (even though I don't play them).
[9:39 AM] beta_helix: It might have been because there weren't enough people trying to work on it, or because they were doing this "on the side" and not as their research projects… but it essentially died out.
[9:40 AM] beta_helix: @spvincent Ok great… I was worried you were sick of design puzzles!
[9:40 AM] jeff101:reposting the ligand puzzle 1855 would be nice. the last time it was up, there was a problem with sharing solutions with teammates. if that problem is fixed, we could do much better on 1855.
[9:40 AM] spvincent: Oh no, they're my favourite.
[9:41 AM] beta_helix: I would love to bring back some ED puzzles, but it is very difficult to find density for natives that have yet to be solved or haven't been publicly released.
[9:42 AM] spvincent: And do you have an endpoint in mind for the covid design binders?
[9:42 AM] beta_helix: horowsah and I recently submitted an NSF grant to try to make Foldit ED waaaaaaaaay better… so we really hope that gets funded and we can hire a developer that will be dedicated to that project.
[9:44 AM] Dudit: @beta_helix: what is the similarity and the difference between Rosetta@Home and Foldit? Can they both be combined?
[9:44 AM] jeff101:I have more questions about hydrogen bonds. Say a certain hydrogen bond is worth 5 points in the Bonding subscore when it is from one segment to another.
[9:45 AM] beta_helix: Great question about covid. We'd have to check with bkoep and the Baker Lab, but I would guess that we will continue as long as there is no treatment available yet. Even then, these puzzles have taught us so much about how to improve Foldit for future binding puzzles (coronavirus or not). If you remember our Flu puzzles from ~10 years ago… my how far Foldit has come (both the game and all of you!)
[9:45 AM] jeff101:If similar strength bond is from one part of the ligand to another, should the Ligand's Bonding subscore also be 5?
[9:46 AM] spvincent: ok, tx
[9:46 AM] jeff101:What if the same strength hbond was between a segment on one monomer to a segment on another monomer? Would both segments have the Bonding subscore 5 ?
[9:47 AM] beta_helix: @Dudit Great question! Rosetta@home is really "donate your CPUs to Rosetta" (you let it use your computer when you are not using it) whereas Foldit is "donate your brain" as YOU are guiding the computations with your folding skills!
[9:48 AM] jmbrownlee333: chuckle
[9:48 AM] jeff101:do groups besides the Baker Lab submit jobs to Rosetta@Home?
[9:49 AM] beta_helix: Yes: https://robetta.bakerlab.org/
[9:50 AM] beta_helix: Anyone can submit a job to Robetta (the robotic Rosetta server… get it? pretty clever, right? and the jobs will be launched to Rosetta@home
[9:51 AM] beta_helix: @Dudit To address your second question…. I'd like to flip that back to all of you. How would you use R@H if we were magically able to incorporate it into Foldit?(edited)
[9:52 AM] spvincent: I've wondered where the name Robetta came from
[9:52 AM] jeff101:does Rosie send jobs to R@H ?
[9:52 AM] :jeff101:https://rosie.rosettacommons.org/
[9:53 AM] Dudit: @beta_helix: I think they should bring back Beginner Coronavirus puzzle, because it will attract more new player (including myself back then)
[9:53 AM] beta_helix: @jeff101 about your subscore 5 question… I would assume that those would act in the same way, but I will check that as well.
[9:53 AM] jeff101:I'm worried about factors of 2. Like if a bond is between 2 segments, both segments have nonzero Bonding subscores that include the 5 points from the h-bond.
[9:53 AM] beta_helix: @Dudit Thanks for the suggestion. Do you think you would still join a new project like this at this point in the pandemic?
[9:54 AM] jeff101:I'm wondering if it is different if the h-bond occurs between 2 monomers or from a ligand to itself.
[9:54 AM] beta_helix: Looking at the Rosie paper, I believe they use their own server at JHU: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0063906
Serverification of Molecular Modeling Applications: The Rosetta Onl…
The Rosetta molecular modeling software package provides experimentally tested and rapidly evolving tools for the 3D structure prediction and high-resolution design of proteins, nucleic acids, and a growing number of non-natural polymers. Despite its free availability to acade…

[9:56 AM] Dudit: @beta_helix: I will do anything that I can to help regarding Coronavirus treatment!
[9:57 AM] jeff101:sometimes I hear about mutants of COVID. Maybe if you had a 3rd type of COVID puzzle, it would attract more players. Are there no other targets we could design binders for?
[9:58 AM] Dudit: There is also COVID 'Bradykinin Storm'
[9:58 AM] jeff101:also, if one of our designs shows promise, I think publicity about it will attract new players
[9:58 AM] jmbrownlee333: Protease inhibitor seem like a good target
[9:59 AM] beta_helix: Before we wrap up, I wanted to share my talk from the Bioinformatics conference that Seth and I presented at virtually over the summer. The special session was called "Bioinformatics outside the lab: How to mobilize online citizen scientists to accelerate research" https://www.iscb.org/ismb2020-program/special-sessions#sst02 and they just posted a video here: https://youtu.be/e8P1EH4WmUY?t=1
YouTube
ISCB
SS02-VIII: Gamers and experimentalists collaborate… - Firas Khati…

Special Sessions
ISCB - International Society for Computational Biology
[10:00 AM] beta_helix: I completely agree with your last 5 comments!
[10:00 AM] beta_helix: Thank you all for stopping by today (I now have to go feed my daughter during her lunch break)
[10:00 AM] beta_helix: Please give us your feedback about these Office Hours: https://fold.it/portal/node/2010431
[10:01 AM] jeff101:thanks for coming and thanks for having these office hours
[10:01 AM] Dudit: thanks @beta_helix:
[10:01 AM] beta_helix: Thank you all for everything that you do!
[10:01 AM] beta_helix: Keep up the great folding…

[agcohn821]

10/9/20- led by horowsah

[1:30 PM] horowsah: Hi all- I'm checking in for an hour of questions if you have any. No worries if you don't, but let me know. I'm not so much an expert on COVID-19 or protein design; my expertise is in electron density type stuff, education, and RNA and DNA and how they interact with proteins, especially during folding.
[1:32 PM] Enzyme2: IMAGE: http://fold.it/portal/files/chatimg/irc_104378_1602271928.png
[1:33 PM] LociOiling: IMAGE: http://fold.it/portal/files/chatimg/irc_476462_1602272012.png
[1:33 PM] LociOiling: IMAGE: http://fold.it/portal/files/chatimg/irc_476462_1602272021.png
[1:34 PM] LociOiling: IMAGE: http://fold.it/portal/files/chatimg/irc_476462_1602272056.png
[1:34 PM] Susume: hi horowsah! at one time there was talk of finding proteins published in the pdb where maybe the match to the ED was not so great, or other metrics were not so great, and having foldit players try to generate a better solution - is that still something that might happen?
[1:35 PM] horowsah: Good question, that is something we're still interested in doing and something that Foldit players can do to help, but sometimes it gets pushed to the back-burner for a bit for things we think will have more immediate scientific impact. However, we will hopefully have more of those in the not too distant future.
[1:38 PM] Susume: I think 'beating' professional scientists is very motivating to foldit players, but I'm not sure how the scientists would feel about it if they are not part of the foldit team - is there any concern about that if we target previously published proteins?
[1:39 PM] horowsah: Yes, I've thought about that quite a bit. I think most scientists are open to others helping out, so long as there's no downside for them. So it's all about how you approach the scientists with the help, even if it happens after Foldit players have worked on it. This happens between scientists as well, so it's something we're used to.
[1:40 PM] horowsah: It doesn't always work perfectly, but I don't think disasters are likely either
[1:40 PM] spvincent: I believe jeff asked a question in feedback about why the sulphur in cysteine couldn't act as as HBond acceptor: I'd be interested in hearing about that too.
[1:41 PM] horowsah: I haven't reviewed that specific case in awhile, so some of my specifics might be slightly off, but it has to do with the details of the electron lone pairs that sulfur has
[1:42 PM] horowsah: They aren't great for accepting hydrogen bonds. Technically, some would say that although they can have hydrogen-bond like interactions, they actually don't meet all of the technical criteria of a hydrogen bond
[1:42 PM] spvincent: tx
[1:42 PM] horowsah: Hydrogen bonding is something of a continuum that we've drawn lines around for what is and what isn't a hydrogen bond, and cysteine falls just outside of it
[1:43 PM] Formula350: I have 2 questions, one that I"m relaying for someone, and then my own. But I'll start with her's :P JoannaH wanted to ask if the Plasmid sequence, the Promoter, and Excretion Tag used for our proteins and both Binder puzzles, could be provided in a downloadable SnapGene format? So that she can try to express and test her protein designs at home.
[1:43 PM] jmbrownlee333: The strength od those H-bonds have been measured and they are quite weak, though not zero.
[1:44 PM] horowsah: On JoannaH's question don't know… but I know who to ask.
[1:45 PM] horowsah: @jmbrownlee333: yes, they are weak. Weak interactions can be important, but they are considered "less important" in most scenarios
[1:46 PM] Formula350: I'll let her know. As for my question (which assumes you occasionally look at player designs): What is the one biggest thing you see that we do in our designs that either: you wish we would stop doing -or- you thing is really interesting and would like more of us to do? (your choice)
[1:47 PM] Jumper2: Or both would be good too
[1:47 PM] spvincent: Is there any issue with having tryptophans on the outside of a monomer in these trimer puzzles? I know they're normally supposed to go in the interior of the protein, but with only one donor they're highly convenient for making networks fully satisfied.
[1:47 PM] Formula350: (I mean, yea, but I didn't want to hog him haha)
[1:49 PM] horowsah: It's interesting, since I'm not an expert in protein design, but more in natural proteins, I'm often surprised at the strategies that are used in protein design, because sometimes they are rather different than what's found in nature. I seem to recall looking at a really good binder the other day and being surprised at how many methionines there were at the interface. It's not something natural proteins do very often, but it seems to work well in design. So my answer really isn't all that handy from helping you to improve, but by trying things that go away from what I expect, I learn new things too. I think others on the team would say that in general, we want Foldit players, and especially experienced Foldit players to try new and different things to help us imagine what's possible that nobody's thought of yet.
[1:51 PM] horowsah: Tryptophans… I'm not familiar with all the details on the trimer puzzles, but tryptophans can be a bit deceiving. There are occasionally natural proteins with exposed tryptophans, but they're pretty rare. From an evolutionary perspective, tryptophan actually is the hardest amino acid to change, which is a bit surprising that it outranks things like glycine and proline. So while it seems like it should be used in a lot of circumstances, nature thinks it's only good for very specific spots, and then never tries to change it.
[1:52 PM] horowsah: Not sure if that actually answers the question, but that's the thing that comes to mind.
[1:53 PM] spvincent: interesting: tx
[1:53 PM] Formula350: I wouldn't be surprised if that design was mine. I'd been trying a couple MET-based solutions. Thanks for that, though (even if it wasn't mine) as I that's exactly the encouragement I wanted to hear; think outside the box.
[1:53 PM] Susume: cool! does nature use tryp in binding interfaces much, or only inside the monomer?
[1:54 PM] horowsah: In my experience, tryptophan is usually internal, except when binding a nucleic acid or nucleotide, where it can stack great with purines. There's probably other examples as well, but those are the only ones coming to mind.
[1:55 PM] Formula350: As an add-on to Susume's, do they often use the Tryp to make HBonds, or are they utilized for their Pi Bonding?
[1:56 PM] horowsah: The h-bond is handy as a way to give specificity, but from what I've seen the stacking is used more when binding other things. When it's a monomer, it's harder to tell, as it's nearly impossible to change one without changing the other in most cases. That might be part of why it's so hard to change via natural evolution.
[1:58 PM] horowsah: I have a question for all of you, actually. When you use the auto-mutate function in a design puzzle, how often do you go in afterwards and change stuff vs accepting what it gives you?
[1:58 PM] jmbrownlee333: for myself, id say almost never.
[1:59 PM] Enzyme2: immediately change them manually. For me rarely. Usually I freeze it if I don't want it to change
[1:59 PM] Formula350: As a hand folder, myself, I do it quite often. A lot of times it's because the Auto-Mutate doesn't "see" the possibilities, and doesn't place in (even at low low CI) what could work well. Where at low CI it likes to inject Phobics instead.
[1:59 PM] spvincent: rarely. Prolines have to be changed manually and sometimes its nice to use asparagine to fix up a Bun
[2:00 PM] jmbrownlee333: I do like to run mutate_no_wiggle occasionally
[2:00 PM] Susume: I watch for places where it put a hydrophilic in the core and change those by hand, and try to figure out how to stop it making the same change again in that spot
[2:01 PM] Formula350: That's why I'm hoping for a feature I'd requested to be approved: White/Blacklisting of AAs in the menu (like how we're blocked from even using them), to make it so the Auto-Mutate tool won't use them. To save myself time from manually brute-force checking all the kinds I'm wanting to try. lol
[2:01 PM] Jumper2: Depends on circumstances, I try to use it to stabilize a protein that I'm trying to get into shape near the beginning. So if you get it into a close position, but afraid it might fly apart if you wiggle, I find the auto mutate at that point to be nice. Then I might hand mutate some to further guide where I want it to go
[2:02 PM] HuubR: I normally watch what Mutate gives me, and when I do not like it (Buns!) I stop it and go one or two steps back, freeze that sidechain, and restart Mutate
[2:02 PM] MikeCassidy: When I have bad loops I manually change them sometines
[2:02 PM] horowsah: @Formula350, that's a feature we've been talking about; not sure if/when it will make it in, though.
[2:02 PM] Formula350: I do as HuubR as well.
[2:02 PM] Formula350: I'll keep my fingers crossed, then
[2:03 PM] jmbrownlee333: I have a random question. why do we start with poly ile
[2:03 PM] horowsah: Ha! I don't know, but let me ask.
[2:04 PM] Susume: sometimes I will manually change all surface proteins to glu or some other blue and see what mutate tool does from there, as a way to jump to a new set of AAs for the same bakcbone
[2:05 PM] MikeCassidy: i also do a 'semi' hand change: I run amutate script on one seg I dont like and see the highest score that isnt what is there then manually change that
[2:05 PM] horowsah: thanks on the thoughts about hand vs auto mutate. I've been thinking about it frequently lately, and wasn't sure how people actually used it.
[2:05 PM] HuubR: My guess would be: Isoleucine is an average size, and no polar atoms on the sidechain to stick to places where you don't want it to, during initial folding. It works fine for me :-)
[2:06 PM] horowsah: Probably average size is a good reason; but I could imagine other choices as well that would work.
[2:06 PM] Susume: they used to give us all alanine, but then our backbones ended up too close together
[2:06 PM] jmbrownlee333: I think its kinda bulky, I think poly val might be an inyteresting twist.
[2:06 PM] horowsah: Yeah, that makes sense, and would mostly be a size thing
[2:07 PM] HuubR: I used VAL earlier on, but I came back from that because the backbone gets too tight for mutate (later on) to have enough choices.
[2:09 PM] jmbrownlee333: I have an education question. Has folditEducation been well received? and is there any community of teachers built around it that you know of? Just FYI I have not used it in the classroom.
[2:10 PM] horowsah: We don't require educators to tell us when they use it, so we don't know completely, but I know of at least ~30 educators using it right now. I'm assuming there are many more who are using it but haven't told us. So far, we haven't had any major complaints, but I expect we'll get more feedback in the next month or so.
[2:12 PM] jmbrownlee333: At what education level do you think its most effective?
[2:12 PM] malphis: I use auto-mutate quite a bit to stabilize before wiggling. Otherwise I'll use a script like Local Mutate. I rarely change individual sidechains by hand.
[2:13 PM] horowsah: We designed Edu Mode to be most complementary to an undergrad biochem class, but I think it would work well in part of a lower level college bio class as well. In the long term, we'd love to make multiple versions aimed at different education levels.
[2:14 PM] HuubR: In your introduction, you spoke about RNA and DNA and how they interact with proteins, especially during folding. I've been wondering whether it makes a difference for folding (in nature!) when you have a helix followed by sheets, or the other way around. Is the first part of the peptide chain already folding while the tail end is still "being assembled"?
[2:15 PM] horowsah: Yes, typically as proteins are synthesized, they start folding before the whole chain is done. So the order is important for having it fold properly in a natural environment. It helps that the ribosome is in itself a pretty good chaperone, and that there are chaperones standing at the ready to help as well, but there are also cases in the literature where you can change the ordering, and it no longer folds properly because it folded in the wrong order while being made.
[2:17 PM] HuubR: So if I design something in Foldit that has a helix at the end, and sheets around that helix preceding it in the sequence, that is not likely to fold like that?
[2:17 PM] Dudit: @horowsah Do we need to understand the protein science to be a good Foldit player?
[2:18 PM] Jumper2: I had thought that might be the case, so on some puzzles, I've tried freezing everything, then slowly unfreezing starting at the first residue. But foldit and I guess rosetta don't seem too well designed to work from an original denatured peptide chain to get to a good fold. It seems you need to get it close to right and foldit will help tweak it an improve it
[2:18 PM] horowsah: @HuubR, it's hard to predict beforehand. Our usual strategy is if it's a stable protein, it will often work fine even if the order of the secondary structures isn't optimal. It's something I'd love for us to improve on predicting.
[2:19 PM] horowsah: @Dudit, that's something I actually don't know the answer to, but I'd be really interested to hear what other vets think.
[2:21 PM] jmbrownlee333: The most successful folders historically dont start out from protein scientist background.
[2:21 PM] Formula350: I continually express that I know nothing about BioChem, and what little I do know, Foldit and the community has taught me. Whether or not I'm a "good folder" (as in, scientifically, if my results are good), I can't say. I'd like to think I am, especially given my limited knowledge
[2:21 PM] Susume: there is that dataset of 4000 foldit designs, I wonder if one could analyze them for short-range vs long-range interactions, how close are the short-range ones to the N-terminal, and does that correlate to in silico success (folding funnel goodness) or invitro success (only have the smaller 145-design dataset for that)
[2:21 PM] HuubR: Depends on what makes a good Foldit player. Playing for points, no, you just need to understand the scoreboard. Playing to help science, yes, some knowledge might help. But on the other hand, too much bias is not good either (as Formula350 is also pointing out :-)
[2:23 PM] jmbrownlee333: I spent more than ten years as a Protein Scientist and I am a middling folder. Thats not data, just anecdote sample size=1.
[2:23 PM] horowsah: I'm going to bring the question of the ordering up. If you're curious, the technical term used typically is called "co-translational folding". Translation is the process of the protein being synthesized, so the term refers to the folding starting before it's done with synthesis. I don't know how well we handle it now, but I'm assuming there are ways we can investigate it, as Susume mentions.
[2:25 PM] jmbrownlee333: It probably helps that we are designing small proteins. which might have smaller problems folding
[2:25 PM] horowsah: I should also mention that I am a biochemist, and I'm not very good at Foldit. I often get paralyzed by trying to fit things into how I've been taught it "shoud" work.
[2:26 PM] jeff101: for tryptophan pi-stacking, which part matters more, the 5-membered ring or the 6-membered ring?
[2:27 PM] Formula350: @Dudit I think having a good understanding of certain things that Foldit doesn;'t directly teach you, is very good (ie Pi Stacking). Knowing very little though, helps, as it lets you think outside the box easier. There's also a lot of ""stupid stuff"" I've done when designing a protein, that later on (say, last week lol) I come to find out, it's a legitimate motif! https://en.wikipedia.org/wiki/Protein_tandem_repeats
Protein tandem repeats
An array of protein tandem repeats is defined as several (at least two) adjacent copies having the same or similar sequence motifs. These periodic sequences are generated by internal duplications in both coding and non-coding genomic sequences. Repetitive units of protein tand…
[2:27 PM] horowsah: Not speaking from an energy perspective but from what I'm seen more typically, the 6-membered ring. But I could be wrong on that one.
[2:27 PM] jeff101: does it use the 6-membered ring to stack with phe and tyr? does it use the 5-membered right to pi-stack with his?
[2:27 PM] Formula350: (in game chatters: that's Protein_tandem_repeats)
[2:27 PM] horowsah: If you're curious, tandem repeats tend to be really good for interacting with other proteins
[2:28 PM] jeff101: do the rings all tend to line up parallel to each other like a stack of pancakes?
[2:28 PM] jeff101: are their centers generally lined up, or is it better to slightly shift their centers from each other?
[2:28 PM] horowsah: pi stacking can have other arrangements, but usually the most favorable would be pancake-style
[2:28 PM] horowsah: Typically, centers are close, but depending on the exact stackers, they can be offset some
[2:29 PM] horowsah: I have to go in a couple minutes, but any last minute questions?
[2:30 PM] HuubR: In the Electron Density puzzle last June (1847), I noticed that at the perimeter of the cloud, there are some parts missing, and other parts seem to be left over from adjacent molecules. Is that due to the fact that the Electron Density map is measured on a crystalline form of the protein, i.e. with a repeating pattern? And if so, how do you know where to cut this pattern to obtain the map for just that one monomer?
[2:30 PM] jeff101: do you have your class playing a contest on Foldit? what should we do when students ask questions on Global Chat?
[2:31 PM] horowsah: @HuubR Yep, that's the basic reason, and it can be really hard at the beginning. Often, it's trial and error for awhile. There are some computational tools that can help, but often they screw it up too.
[2:31 PM] jeff101: are there ways Foldit can block Global Chat so students in certain classes cannot use Global Chat?
[2:32 PM] horowsah: @jeff101: Yes, I and others do have students playing contests. There are ways we can block it if the educator wants to, but we've typically left it up to them. This has been something we've debated about how to deal with best.
[2:32 PM] jeff101: as a teacher, you might want to block your students from using Global Chat so they can't get help from other Foldit players
[2:32 PM] MikeCassidy: Yes I am always wondering if I am helping students too much BUT I dont always know if it is a student or someone like myself just playing
[2:32 PM] Skippysk8sIRC: or use us as a cheap teaching assistant lol
[2:33 PM] HuubR: jeff101, this reminds me of a post on the forum:
[2:33 PM] HuubR: https://fold.it/portal/node/2009529
[2:33 PM] Formula350: Yes, I tried a Tandem Repeat this last time on the Spike puzzle! It… didn't play out so well, but I didn't get to devote as much time to it. I named it the "Triple Deuce" (a musclecar term).
[2:33 PM] horowsah: Personally, I don't mind if they get help from other Foldit players, because they still have to actually do it themselves. My primary concern is overtaxing our players with a lot of questions that their instructor should be answering instead.
[2:33 PM] Skippysk8sIRC: we can always need to answer the phone ;)
[2:34 PM] jeff101: it would help to have videos of solving the Edu puzzles if they differ from the usual Intro Puzzles
[2:34 PM] horowsah: It probably would be… something to think about.
[2:35 PM] horowsah: I have to head to another meeting, thanks everyone!
[2:35 PM] Formula350: I've personally abstained from providing too much help to students so that they can still "discover" the answer/solution. Personally I've found that being given the answer hasn't exactly helped my learning heh Though, certaintly there are times it has, so it's tough to say
[2:35 PM] Skippysk8sIRC: thanks
[2:35 PM] Formula350: Thanks Horowsah!
[2:35 PM] Dudit: thank you @horowsah
[2:35 PM] jeff101: yes. thanks for coming. I enjoy these office hours.
[2:36 PM] HuubR: Thank you, horowsah!
[2:36 PM] Susume: thanks horowsah!
[2:36 PM] MikeCassidy: Thanks for the talk

[agcohn821]

10:59 AM] bkoep: Okay, office hours are open!
[10:59 AM] donuts554: hello
[10:59 AM] bkoep: I'm a scientist at the UW Institute for Protein design, and I oversee a lot of Foldit's protein design puzzles. Happy to take any questions about Foldit and/or protein design!
[10:59 AM] formula350: Hello Sir
[10:59 AM] BletchleyPark: Hello bkoep
[11:00 AM] donuts554: hello how are you?
[11:00 AM] Dudit: Hi Bkoep
[11:00 AM] alcor29: Any news on LCB1 trials?
[11:01 AM] BletchleyPark: Is there news on the energy landscape publication ?
[11:01 AM] Dudit: Is it possible to design any protein that didn't require any energy at all?
[11:01 AM] bkoep: Any news on LCB1 trials? @alcor29 No, the last I heard was preparations for experiments in mice, and maybe non-human primates
[11:01 AM] alcor29: Tx bkoep.
[11:03 AM] bkoep: BletchleyPark: Is there news on the energy landscape publication ?
It has not yet been accepted by a journal for publication. We recently heard back from the editor who wanted some revisions (normal for any paper)
[11:04 AM] BletchleyPark: ok, thank you. Is it possible to ass the supplmental list of foldit players to the preliminary publication ?
[11:04 AM] BletchleyPark: to add
[11:04 AM] pc: :D
[11:04 AM] bkoep: Is it possible to design any protein that didn't require any energy at all? @Dudit Can you elaborate what you mean?
[11:05 AM] pc: is it possible to create variant of LCB1 protein that work in lab, or when we change one AA, its moslty doesnt work anymore ?
[11:06 AM] MikeCassidytoo: hi all
[11:06 AM] donuts554: hello how are you?
[11:07 AM] Dudit: A working and folding protein that is using no energy, because the current design is still using energy (maybe thermodynamic factor)
[11:07 AM] bkoep: BletchleyPark: ok, thank you. Is it possible to ass the supplmental list of foldit players to the preliminary publication ?
I'm not sure if we will be able to revise the authors. But if you played the old monomer design puzzles and would like to be added, it wouldn't hurt to fill out the authorship form
[11:07 AM] bkoep:

docs.google.com

Energy Landscape Optimization Paper

This form is open to all Foldit players who have played a "Monomer Design" puzzle in Foldit (like Puzzle 1698: Medium Monomer Design). Researchers at the Baker Lab have used Foldit players' work to help develop a new method for designing proteins....

Google Docs
Energy Landscape Optimization Paper
This form is open to all Foldit players who have played a "Monomer Design" puzzle in Foldit (like Puzzle 1698: Medium Monomer Design). Researchers at the Baker Lab have used Foldit players' work to help develop a new method for designing proteins. They are preparing a researc…

[11:08 AM] BletchleyPark: I played the puzzles and I received the authors list, but that list is not part of the current online publication, hence we're not credited as individuals, other than 'foldit players '
[11:10 AM] bkoep: is it possible to create variant of LCB1 protein that work in lab, or when we change one AA, its moslty doesnt work anymore ?
@pc: I'm not sure what you're getting at. But the original LCB1 paper described the effects of mutations to LCB1 (fig 2)
[11:10 AM] bkoep:
https://science.sciencemag.org/content/early/2020/09/08/science.abd9909
De novo design of picomolar SARS-CoV-2 miniprotein inhibitors
Targeting the interaction between the SARS-CoV-2 Spike protein and the human ACE2 receptor is a promising therapeutic strategy. We designed inhibitors using two de novo design approaches. Computer generated scaffolds were either built around an ACE2 helix that interacts with t…

[11:10 AM] pc: thanks
[11:10 AM] bkoep: (I apologize if that article is behind a paywall)
[11:11 AM] susume:
https://www.bakerlab.org/index.php/publications/
Baker Lab
admin
Publications - Baker Lab
To read an article, click on its title and select the PDF file. Use the box below to search publications.
[11:12 AM] susume: you can click on article title here and get free pdf
[11:12 AM] bkoep: A working and folding protein that is using no energy, because the current design is still using energy (maybe thermodynamic factor)
@Dudit Protein folding is passive for all of our protein designs. The folded state has a lower free energy than the unfolded state
[11:13 AM] bkoep: We have a deeper discussion about free energy and protein folding on the blog
[11:13 AM] bkoep:
https://fold.it/portal/node/2005623
The problem of protein design
The problem of protein design
[11:14 AM] formula350: Clickable:
[11:14 AM] formula350: https://fold.it/portal/node/2005623
The problem of protein design
The problem of protein design
[11:14 AM] donuts554: I thought that a string of glutamatic acid residue didnt need any energy to fold
[11:14 AM] alcor29: On the two sided interface puzzle. Do the two helixes have to exactly that far apart? Could they have been a bit closer to each other?
[11:15 AM] bkoep: BletchleyPark: I played the puzzles and I received the authors list, but that list is not part of the current online publication, hence we're not credited as individuals, other than 'foldit players '
That's right, the pre-print was posted early, before players had much chance to fill out the authorship form. The pre-print does not include individual Foldit player names, but any final publication will include all names – probably in the supplementary info.
[11:15 AM] jeff101: have you ordered any Foldit-designed binders lately?
[11:15 AM] Dudit: How many % probability of successful working protein design from Foldit player in the wet lab if the new metrics (DDG,SASA,SC) is applied?
[11:16 AM] susume: is there a comment area on the preprint where you could link to a page listing individual authors?
[11:16 AM] bkoep: On the two sided interface puzzle. Do the two helixes have to exactly that far apart? Could they have been a bit closer to each other?
@alcor29 Yes, we're interested in an interface that is compatible with that exact distance between the two hairpins
[11:17 AM] bkoep: This is a general problem in protein design, and one that our computer algorithms are particularly poor at. So this two-sided interface puzzle is an exciting pilot experiment for us.
[11:17 AM] bkoep: Depending on the results, we may see a lot more like it
[11:18 AM] alcor29: On DDG. Am able to get the metrics spot on but it means going contrary to my intuition. Does that interfere with the human contribution to protein folding design.
[11:19 AM] bkoep: jeff101: have you ordered any Foldit-designed binders lately?
No, we haven't ordered any more designed binders for lab testing. We know that we need to work on the properties of our designs in Foldit. We won't want to run another lab experiment until we have the new Metrics in place and have more designs that satisfy our Metric thresholds
[11:21 AM] donuts554: Yes the Metrics are important
[11:21 AM] bkoep: How many % probability of successful working protein design from Foldit player in the wet lab if the new metrics (DDG,SASA,SC) is applied?
@Dudit It's hard to say for certain. But LCB1 was selected with very similar metrics, as part of a pool of 100,000 designs. They found about 100 hits in that experiment, so that's a success rate of 0.1%.
[11:21 AM] Dudit: I think there should be a new list in the Achievement page called 'Successful Working Solution' when any Foldit player protein design works successfully in the wet lab
[11:22 AM] BletchleyPark: What is the real-world purpose of the two-sided interface problem ? and there is still an open question from Susume
[11:23 AM] jeff101: if a segment is involved in pi-stacking, what subscores will best reflect this?
[11:24 AM] jeff101: do only trp, phe, tyr, and his do pi-stacking? can arg and pro do it too?
[11:24 AM] bkoep: On DDG. Am able to get the metrics spot on but it means going contrary to my intuition. Does that interfere with the human contribution to protein folding design.
@alcor29 Not at all! There will always be some amount of learning for humans to play Foldit and contribute to protein design. There are also matters of presentation and communication – we may be able to improve how DDG is presented in Foldit, in a way that makes it more intuitive to address.
[11:25 AM] donuts554: I dont think arg and pro can do it because they are not aromatic
[11:25 AM] HuubR: About "designs that satisfy our Metric thresholds": do you mean that, for instance, the DDG has to be -40 or better, or otherwise the solution will not be viable in the wet lab?
[11:25 AM] alcor29: k
[11:27 AM] Skippysk8sIRC: donuts, pro isn't an aromatic but it has a loop. arg is totally different
[11:27 AM] formula350: Donuts: Arg can Stack, and funny Jeff mention Proline, as I came across an instance in the Aflatoxin the day before where it certain looked like the PHE and PRO were stacking (based on Sphere view)
[11:27 AM] formula350: certainly*
[11:28 AM] bkoep: is there a comment area on the preprint where you could link to a page listing individual authors?
@Susume Maybe… although I'm not sure if that would meet the comment policy of biorXiv
[11:28 AM] alcor29: Is there any way of introducting more flexibility in structures. For exmample, to satisfy the new metrics it might help if the sss could bend a little?
[11:29 AM] bkoep: I think there should be a new list in the Achievement page called 'Successful Working Solution' when any Foldit player protein design works successfully in the wet lab
@Dudit I like this idea, but we are not set up very well to support it right now
[11:29 AM] alcor29: Might be a par tof the wiggle function. Or the rigidity may just be an artifact of the graphic representation.
[11:32 AM] Skippysk8sIRC: to Alcor's question, does tighter binding of side chains or SS make the shape conform better?
[11:32 AM] bkoep: BletchleyPark: What is the real-world purpose of the two-sided interface problem ? and there is still an open question from Susume
This particular puzzle would help with an IPD project about modular protein complexes. If we decide to run more puzzles like it, then I think the project leader would write up a full blog post with details about the project
[11:32 AM] BletchleyPark: ok, thank you
[11:33 AM] pc: this should be on puzzle description ^^
[11:33 AM] BletchleyPark: Which supercomputer did you use to generate the big set of proteins for covid ?
[11:33 AM] donuts554: I have 3 questions
[11:34 AM] alcor29: Thanks Skippy. Good addition.
[11:34 AM] bkoep:, jeff101: if a segment is involved in pi-stacking, what subscores will best reflect this?
Rosetta/Foldit actually does not have a term for pi-stacking. So the energy associated with complementary pi-orbitals is not included in Rosetta calculations. This may change in the future, but for now, stacked aromatics should still have a good Packing subscore.
[11:35 AM] donuts554: One is what is the purpose of the dense concentration of oxygen atoms in front of the locked receptor part in puzzle 1855?
[11:35 AM] donuts554: https://fold.it/portal/files/chatimg/irc_902783_1593572159.png
[11:36 AM] formula350: (I've observed very good Packing in my Pi Stacked instance, for whatever that's worth for ya, Jeff)
[11:36 AM] donuts554: Two is are these sheets and why or why not?
[11:36 AM] donuts554: IMAGE: http://fold.t/portal/files/chatimg/irc_902783_1602956212.png
[11:37 AM] Dudit: I think there should be a Bradykinin Storm Coronavirus puzzle in Foldit
[11:37 AM] formula350: (Highest, albeit in the middle of a Trimer, was over 200 on that Subscore, between PHE)
[11:37 AM] bkoep:
@HuubR: About "designs that satisfy our Metric thresholds": do you mean that, for instance, the DDG has to be -40 or better, or otherwise the solution will not be viable in the wet lab?
I mean that we have evidence that these Metric thresholds improve success rates. Even when we meet these thresholds, our success rate is somewhere around 0.1%. If we don't meet the Metric thresholds, we expect our success rate will be less than that.
[11:38 AM] HuubR: Thanks
[11:40 AM] pc: oh that means we realy need a very good ddg
[11:40 AM] donuts554: And three is that is a design like this stable, and why or why not?
[11:40 AM] donuts554: https://fold.it/portal/files/chatimg/irc_902783_1596412421.png
[11:40 AM] donuts554: https://fold.it/portal/files/chatimg/irc_902783_1596414166.png
[11:41 AM] pc: 1500 sasa and -40 ddg are hard to reach in design puzzles. We are often in 1300 sasa and -35 ddg in mostly top solutions
[11:41 AM] Skippysk8sIRC: so back to alcor's question perhaps?
[11:43 AM] bkoep: Is there any way of introducting more flexibility in structures. For exmample, to satisfy the new metrics it might help if the sss could bend a little?
@alcor29 This is a little bit dangerous, and we should be careful about over-optimizing with the metrics. The metrics are useful for evaluating whether a protein interface looks good. But we need to be very cautious about how much alter our models to make them conform to the Metrics. You might be able to get better SASA if you bend your helices, but bent helices can also make your protein design unrealistic and unlikely to fold
[11:44 AM] bkoep: There is perhaps an element of Goodhart's Law in that concern with the metrics
[11:44 AM] HuubR: "When a measure becomes a target, it ceases to be a good measure."
[11:45 AM] HuubR: (had to look that up :-)
[11:45 AM] bkoep: BletchleyPark: Which supercomputer did you use to generate the big set of proteins for covid ?
The UW has a large computer cluster that we can use, and we also have a pretty nice cluster at the IPD itself
[11:46 AM] pc: If the measure need to resolve lot of difficult constraints to have hight score, it can be a good measure(edited)
[11:47 AM] BletchleyPark: ok, thanks, I thought you had used ORNL
[11:48 AM] bkoep: donuts554: One is what is the purpose of the dense concentration of oxygen atoms in front of the locked receptor part in puzzle 1855?
I'm not sure! I don't know a lot about the Y1 receptor in Puzzle 1855, but those oxygens could be important for how the Y1 protein associates with other proteins
[11:48 AM] BletchleyPark: (my suggestion to Baker Lab last year)
[11:48 AM] Skippysk8sIRC: how much might a real protein shape change when it binds to something? We know that the Covid spikes have hinges and move. Is this atypical?
[11:49 AM] formula350: @bkoep Something I've been trying to ask at Office Hours lately: What is one of the most common things you see us doing with our designs that either: you wish we would cut back on -or- you think is really interesting and would like it if more of us did it? (your choice; answering both is always welcomed lol)
11:49 AM] bkoep: donuts554: Two is are these sheets and why or why not? donuts554: IMAGE: http://fold.it/portal/files/chatimg/irc_902783_1602956212.png
I'm worried there is not enough hydrogen bonding between those sheets. Beta sheets should form a ladder of H-bonds between adjacent strands
[11:49 AM] formula350: (Skippy seems to be asking about Conformal Change (something JoannaH is hoping might be a thing Foldit can incorporate/explore)
[11:50 AM] bkoep: donuts554: And three is that is a design like this stable, and why or why not? donuts554: https://fold.it/portal/files/chatimg/irc_902783_1596412421.png
It's hard to tell from the images, but I'm concerned that this protein lacks a well-packed hydrophobic core.
[11:54 AM] BletchleyPark: Can foldit be compiled for 64-bit use as well to overcome size limitations ?
[11:55 AM] jeff101: @bkoep, thanks for having this Office Hour and for all your answers so far
[11:55 AM] bkoep: Skippysk8sIRC: how much might a real protein shape change when it binds to something? We know that the Covid spikes have hinges and move. Is this atypical?
This is a tricky subject. In natural systems, it is common for proteins to change shape upon binding (you can probably find more resources online about "induced fit"). However, there is usually some energy cost associated with that change in shape (otherwise, the protein would have take that shape from the start). If you can design a folded protein that is pre-organized in exactly the shape in needs to bind, then you can avoid that energy cost and improve the binding affinity.(edited)
[11:55 AM] Jumper2: I would expect that handling conformal changes would be rather difficult to add into the code given that it turns a puzzle session into multiple parallel sessions each trying to get to a different solution. If I had to try it today with Foldit, I'd probably try to set up each conformation as a track, possibly bring in the alignment tool. On a similar note, has anyone noticed if the set of Foldit players solutions for a particular puzzle end up finding multiple conformations?
[11:56 AM] jeff101: the Partition Puzzles from 2019? dealt with that
[11:57 AM] Jumper2: It would be somewhat heartening to know that when looking at the set of all solutions for a puzzle, that we at least manage as a group to statistically cluster around known conformations
[11:58 AM] donuts554: But it does form a ladder of Hydrogen bonds both on the x and z axes
[11:58 AM] donuts554: IMAGE: http://fold.it/portal/files/chatimg/irc_902783_1602957529.png
[11:59 AM] donuts554: IMAGE: http://fold.it/portal/files/chatimg/irc_902783_1602957544.png
[11:59 AM] donuts554: IMAGE: http://fold.it/portal/files/chatimg/irc_902783_1602957552.png
[11:59 AM] BletchleyPark: #binders If we find a binder for Covid, what are the odds it will make it into real medication ?
[12:00 PM] susume: that's very pretty, donuts
[12:00 PM] jeff101: The Partition Function Tournament of 2018:
[12:00 PM] jeff101: https://fold.it/portal/node/2005623
The problem of protein design
The problem of protein design
[12:01 PM] jeff101: https://fold.it/portal/node/2005638
Partition functions
Partition functions
[12:01 PM] jeff101: https://fold.it/portal/node/2005660
Protein Design Partition Tournament
Protein Design Partition Tournament
[12:01 PM] jeff101: https://fold.it/portal/node/2006103
Partition Tournament Final Results
Partition Tournament Final Results
[12:02 PM] formula350: @bkoep Something I've been trying to ask at Office Hours lately: What is one of the most common things you see us doing with our designs that either: you wish we would cut back on -or- you think is really interesting and would like it if more of us did it? (your choice; answering both is always welcomed lol)
In the symmetric trimers, we have to reject a lot of designs because they rely on a tight triangle of H-bonds in the middle of the protein (with TYR, THR, or SER residues). These "networks" score better than they should in Foldit, and we've been trying to figure out a good way to fix it. Ideally, the angle between H-bonds is usually tetrahedral (109 degrees) or trigonal planar (120 degrees); these tight triangles require three close-proximity H-bonds at 60 degree angles.
[12:02 PM] alcor29: On the two sided interface. There seems to be a difficulty in getting rid of non-ideal loops due to the distance. Is that just something we need to overcome, or are non-ideal loops kind of to be expected?
1
[12:02 PM] bkoep: I talked about these triangles a little bit in one of the recent Lab Report videos
[12:03 PM] formula350: That's good to know, thank you :)
[12:04 PM] bkoep: BletchleyPark: Can foldit be compiled for 64-bit use as well to overcome size limitations ?
I'm not certain what size limitations you mean? Our mac version of Foldit is compiled for 64-bit and we haven't noticed any significant differences in performance
[12:04 PM] pc:: yes this kind of informations are very usefull for us ^^ (nice question formula )
1
[12:04 PM] robgee: Bah… those tight triangles are the easiest to make :p
[12:05 PM] pc: why some design puzzles allow threonine serine, and other design puzzles not?
[12:05 PM] LociOilingIRC: skip the triangles…more cowbell
[12:05 PM] jeff101: :)
[12:06 PM] BletchleyPark: #64bit I recall 32-bit windows clients crashing when they memory usage reaches around 1.6 Gb
[12:08 PM] BletchleyPark: The protein design sandbox puzzle will crash eventually
[12:10 PM] bkoep: BletchleyPark: #binders If we find a binder for Covid, what are the odds it will make it into real medication ?
It's a long road from the lab to the pharmacy. Especially because de novo protein design is so new and, to my knowledge, no de novo protein has been approved by the FDA (I think the closest might be Neoleukin's IL-2/15 mimetics, which may start clinical trials soon). So, hopefully we will have something effective for COVID-19 long before our binders have a chance to make it to the shelves
[12:12 PM] BletchleyPark: thank you
[12:12 PM] bkoep: (We are hopeful the path-to-pharmacy may get shorter in the future as more de novo protein drugs are developed. So the next pandemic could be a different story)
[12:13 PM] Skippysk8sIRC: is it getting easier to get AAs for the wet lab now? or is the supply chain still running slow
[12:14 PM] bkoep: On the two sided interface. There seems to be a difficulty in getting rid of non-ideal loops due to the distance. Is that just something we need to overcome, or are non-ideal loops kind of to be expected?
@alcor29 Yes, that's a known bug. The Ideal Loops Objective was not designed to work with chain breaks. I believe it will be impossible to completely satisfy that Objective in the puzzle, but hopefully we can do something to ameliorate the issue in future puzzles
[12:15 PM] HuubR: (Was that a challenge to all of us?)
[12:15 PM] mikelewis: haha I thought the same thing Huubr
[12:16 PM] formula350: @Jumper2 I wonder if it would genuinely be that "hard" to accommodate a Conformation System… I think the Devs technically have all the 'tools' right now that would allow it. As in, allowing certain backbone segments unlocked status, but having Constraint Bans of a weak strength attached (and probably attached to their Bondable atoms) to literally "keep them on a short leash".
[12:17 PM] formula350: That would allow them to move a bit, to conform, but not have totally free movement like we do in our own proteins.
[12:17 PM] bkoep: why some design puzzles allow threonine serine, and other design puzzles not? @pc: Good question! There is some evidence that SER and THR can interfere with how helices fold, so we like to avoid those residues in helices if possible. However, SER and THR are really useful for H-bond Networks and satisfying BUNS, so sometimes we allow them.
[12:18 PM] pc:: thanks
[12:20 PM] alcor29: Thanks for the ov hour plus plus bkoep!
[12:20 PM] mikelewis: Yep, thanks!
[12:20 PM] Skippysk8sIRC: yes, thanks. We look forward to the working binder metrics
[12:20 PM] pc: yes thanks ^^
[12:21 PM] formula350: bows Thank you, Science Senpai, for recognizing us.
[12:21 PM] Dudit: @bkoep thank you very much!
[12:21 PM] pc: dont hesitate to experiment multiple new metrics score system in devprev, we will test them
[12:21 PM] jeff101: Saturday seems like a good time for Office Hours. Seems like many people came today.
[12:22 PM] Skippysk8sIRC: I gave up my second farmer's market for this… maybe lost out on melons
[12:22 PM] BletchleyPark: We usually have jobs too, so yes saturdays are fine, thank you for taking the time today bkoep
[12:22 PM] formula350: Might be due to it being The Man's office hours lol But yes, it seems Sat IS a better day. However, I can appreciate the others not quite wanting to use an hour of their saturday for 'work'.
[12:22 PM] HuubR: Thanks, bkoep
[12:24 PM] jeff101: Yes, thanks bkoep.
[12:24 PM] bkoep: That's good feedback to hear that Saturdays are convenient!
[12:24 PM] bkoep: I think we will still continue to shuffle Office Hours a little bit to try to accommodate different schedules, but maybe we can start including more Saturdays
[12:25 PM] bkoep: Thanks all for the great questions!
[12:25 PM] formula350: So yesterday I came across a curious instance where scientists modified Leucine tho mimic a Methionine without the Sulphur (I presume; it was to 'prevent oxidation'), which this was called Norleucine. I wonder if something like that could be done to Cysteine, too, to give us an extra sidechain to work with… (feel free to field this if you want bkeop lol)
[12:26 PM] BletchleyPark: I'm off now, good night
[12:26 PM] formula350: Cya Bletchley
[12:27 PM] robgee: Bye Bletchley, Thanks bkoep
[12:29 PM] bkoep: @Formula350 We like to stick to the 20 canonical AAs as much as possible. Mainly because they are easy to test (just about any organism in the world can translate a protein made from the 20 AAs). If we started incorporating non-canonical AAs, we would have to use specials modified E. coli to test them, or else synthesize them chemically. That makes testing more difficult.
[12:29 PM] formula350: snaps fingers Drat. :P Thanks though :D
[12:30 PM] Jumper2: After getting into playing Foldit, I've found that I also enjoy just browsing around the PDB database looking at interesting structures. One thing I've noticed is that the so many of the biological proteins have "bad" aspects to them from what might be a score perspective. Just viewing a few, the categories of "outliers" is impressive (Ramachandran, Rotamer, RSRZ, Angle, RSRCC, Mogul-angle, Mogul-bond). And many natural proteins seem loaded with these outliers. So questions like alcor29's about what is an expected "red flag" that could be ignored versus a genuine issue really help us out in trying to get a handle on finding solutions. One thing that might be useful would be to load some sample PDB proteins into Foldit puzzles just so we could see how they show up in the score and various objectives. I think that would go a long way towards improving our qualitative grasp on what things we really need to concentrate on.
[12:32 PM] formula350: Jumper, press keyboard Up Arrow. press CTRL+A, then CTRL+C. Go to #bugs-and-feedback. Press CTRL+V, hit ENTER lol (I support that idea, though)
[12:42 PM] pc: for the Goodhart's Law, yo avoid this problem, banks use simulators that have lot of parameters to make good economic predictions. The Goodhart's Law is not a measure that is too much accurate, it is just that we use too much only one.(edited)
[12:50 PM] bkoep: @Jumper2 That's a great observation about imperfections in PDB structures, and I'll chime in that there are at least two important caveats to consider here: 1. Poor quality models It is not trivial to build a protein model from experimental data (X-ray diffraction, cryoEM, etc.). If the model builders are not very careful, it is easy to over-fit your protein model to noisy experimental data. This can lead to outliers in PDB models that are probably errors, and do not actually reflect the true structure of the protein. 2. Natural proteins are non-ideal We should expect natural proteins to have some non-ideal features – not because they represent good design principles – but because they can get away with it. Natural proteins are not optimized for folding stability, but for organism fitness (via natural selection). So many natural proteins are right on the boundary of folding stability. Furthermore, they are not optimized rationally, but "accidentally" over billions of years of random mutation. For protein design purposes, we are most interested in discovering robust design principles that can be used to rationally design a protein in minutes/hours. This is maybe a reason to be cautious about how much we guide to nature in the field of protein design.(edited)
[12:52 PM] Jumper2: Definitely good information, thanks! I'm guessing #1 is why there's the validation section prominent on the front page (the red to blue bars)
[12:53 PM] pc: yes interesting. In foldit we try to create clean and safe proteins, to make sure they work ^^(edited)
[12:54 PM] formula350: Which I think highlights exactly why people shouldn't go out of their way to look up a Nature made protein on PDB to get its sequence and try to recreate it for a design in Foldit, since your results likely won't produce the intended results. (Or if it does, in Foldit, it stands to reason that it may not in the wet lab)
[12:57 PM] Jumper2: Agree on #2 as well. It seems that one of the big questions we (the planetary we) need to answer is that if we limit our designs to stable, robust constructions that can be relied upon, have we excluded certain types of functionality because those are inherently only possible within the domain we're trying to avoid.
[12:58 PM] Jumper2: An answer to that question would be nice to have sometime in the next half century
[12:59 PM] Jumper2: Thanks for taking the time today to answer our questions!
[1:01 PM] formula350: the extended time, at time

[agcohn821]

6:05 PM] beta_helix: Hi everyone! We'll start today's Foldit Office Hours very soon.
[6:05 PM] formula350:riggety-REMIIIXXX airhorns blare
[6:05 PM] pc: hi
[6:06 PM] beta_helix: Hello everyone. I'm here with Seth Cooper (the co-creator of Foldit) for Foldit Office Hours!
[6:06 PM] sethcooper:hello!
[6:06 PM] formula350:beta_cooper in the house.
[6:06 PM] jeff101: Hi! Thanks for coming
[6:07 PM] beta_helix: Thank you for stopping by
[6:07 PM] Todd6485577: Happy Folding all
[6:08 PM] formula350:Well I guess I'll jump in and ask either one of ya: What's the chances we'll ever get more Building Blocks for Blueprint mode?
[6:09 PM] Todd6485577: Or like customized building blocks, like select segment save as building block?
[6:10 PM] sethcooper:I think we have discussed improving them
[6:10 PM] sethcooper:I'm not sure about saving existing segments thought!
[6:10 PM] sethcooper:that sounds useful
[6:11 PM] beta_helix: We've talked about adding a much wider selection of loops and building blocks…
[6:11 PM] formula350:Yes, I have that in my Feature Request document for the Devs. The ability to save our own blocks, of even larger designed portions. (Only ones that have "passed" filters though)
[6:11 PM] JohnMcLeod: oh joy :-)
[6:11 PM] beta_helix: and hopefully this is something we could get to in the not too distant future.
[6:11 PM] alcor29: 1. What is the thinking behind doing a MERS binder? 2. What is the thinking behind exploring that 2 corona binders linked together is berrter than one?
[6:12 PM] pc: Why there is no sars-cov2 binder design puzzles anymore ? will there be other later ?
[6:12 PM] formula350:Thanks, I'm glad to hear it's been discussed :)
[6:13 PM] beta_helix: @pc I believe the IPD team wants to do more analysis on your sars-cov2 binder design puzzles first, before posting more of them.
[6:13 PM] pc: thanks
[6:13 PM] milkshakeiii: 1912b is technically sars-cov2 right?
[6:13 PM] milkshakeiii: with two spikes
[6:14 PM] formula350:Yes, and LCB1 and LCB3 designs that we're trying to "link" together.
[6:14 PM] beta_helix: Similar to the addition of metrics, first and foremost we want to make sure that all of your effort yields as promising structures as possible
[6:14 PM] jeff101: I think there is something wrong with the LUA commands save.LoadSolution() and save.LoadSolutionByName(). I wrote a Feedback about it:
[6:14 PM]
jeff101: https://fold.it/portal/node/2010726
[6:15 PM] jumper2: A recent puzzle mentioned that loops would be penalized if they didn't match one of the available building blocks. Unfortunately it didn't specify "match" Would a helix-helix block that was Red,Green,Blue,Blue,Red count as match or do the Rama angle have to be within some defined but unpublished epsilon of what the building block uses?
[6:15 PM] zo3xia: I shared a new update. Doesn't seem to freeze foldit
[6:16 PM] pc: Do we know why proteins have more chance to works in lab with SC > 0.6 in design puzzles ?
[6:17 PM] sethcooper:I'm not totally sure why those functions might be missing solutions
[6:18 PM] sethcooper:it looks like they should have a solution at index 0 though
[6:19 PM] beta_helix: @alcor29 (Re 1) The idea with the MERS puzzles is that it would be extremely useful to generate a successful solution to this. Obviously this is the case for any disease, but with MERS it is so similar to covid, that you already have a ton of experience tackling this type of topology.
[6:19 PM] Todd6485577: I would guess because the SC factors into account parts of the overall shape that would be formed once it is bound.
[6:20 PM] alcor29: k
[6:21 PM] Todd6485577: but tat's just my guess
[6:22 PM] beta_helix: @Jumper2 can you tell me which puzzle that was?
[6:22 PM] jumper2: Let me see
[6:25 PM] pc: (For the SC question, it is more why it is 0.6 and what happen in reality)
[6:25 PM] donuts554:Why is there a cluster of oxygen atoms on one side of the locked and unlocked part in the Aflatoxin Challenge puzzle?
[6:25 PM] donuts554:https://fold.it/portal/files/chatimg/irc_902783_1588135313.png

[6:25 PM] Ridick051: what is the plan for the future of foldit? where is the journey going?
[6:27 PM] sethcooper:something we have been thinking about for a while is VR
[6:28 PM] jumper2: for the building block challenge look at 1913 even, it's in the first comment section by bkoep under "Ideal Loops"
[6:29 PM] beta_helix: @alcor29 (Re 2) I believe the spikes that were bound the design from the IPD were trimers: https://science.sciencemag.org/content/sci/370/6515/426/F4.large.jpg

[6:30 PM] jumper2: The SARS-CoV-2 spike protein is definitely a trimer of RBD's. I believe they can open and close independently
[6:30 PM] beta_helix: But rigid linker design is a super tough problem in protein design, so the first step is to work with 2 copies of the spike.
[6:30 PM] formula350:Jumper, I suspect that it's a simple wording snafu. As you notice, on a number of the puzzle's for his Objectives explanations, he actually words the one for SS Design as "Penalizes for CYS", when in fact CYS is always disabled.
[6:30 PM] formula350:(the one time he forgot to post that, I posted it and corrected that, which it looks like he might've copy-pasted that one since 1913 has it correctly worded as "Disabled" heh)
[6:31 PM] beta_helix: We are very exciting about these Designable Linker Puzzles (can I call them DLPs? ) because this is clearly something that computers have a lot of trouble with…(edited)
[6:31 PM] alcor29:Thanks beta_helix:.
[6:31 PM] beta_helix: …and hopefully they are fun to work on!
[6:32 PM] Ridick051: yes it is ^^
[6:32 PM] Skippysk8sIRC: new shiny toy
[6:32 PM] Susume: @Jumper2 I can't remember if loops are evaluated on phi,psi or on rmsd to the building block, but I do know you can be in the right abego colors and not be close enough to the target shape to get credit for the loop
[6:32 PM] formula350:Not us human computers, we'll do to it like we did with the… what was it, Monkey Virus or AIDs, in which we solved in a couple weeks what had stumped scientists for 15 or so years lol
[6:33 PM] beta_helix: The Contenders actually solved it in less than 10 days
[6:33 PM] jumper2: If you're binding to one of the RBD's while it's extended, and the virus can retract them to prevent detection by the immune system until it's in a more favorable environment (temp, pH, etc.) then by using a DLP, do you have to have two extended for it to bond? That seems like it would be worse since any virus with 0 or 1 extended would get a pass through
[6:33 PM] alcor29: DLPs it is!
[6:33 PM] jumper2: Thanks @Susume
[6:34 PM] jumper2: Or if a DLP bound to a virus with one RBD extended, could it possibly jam a second one closed?
[6:35 PM] jumper2: Given the symmetry of the trimer, if you had a single copy of LCB1 or LCB3 to bind to a single extended RBD, then had some "extra junk" hanging off of it that could jam the two other RBD's in the down position
[6:36 PM] beta_helix: Seth, are you an expert on DLPs? I certainly am not, but my guess would actually be that this is a relatively new field and we probably don't have that many cases of them working. (This is totally speculation on my part!).
[6:36 PM] beta_helix: I believe the team was planning on having a specific chat on DLPs (with those who created the puzzle)
[6:37 PM] formula350:I suspect that with how many "Spikes" there are on the whole virus, as well as how much of the DLP would be injected into us, that the chances are good for enough of the domains to be open.
[6:37 PM] jumper2: It just seems from the literature that it might be common to find a virus with only one extended RBD (or maybe that's just how people like to draw the thing)
[6:37 PM] Skippysk8sIRC: applause
[6:37 PM] beta_helix: Seth, does that sound right? If not, we should schedule that anyway!
[6:37 PM] formula350:There's a blog post about the DLP, it's just a primer though.
[6:37 PM] sethcooper:yeah it seems worth having a specific chat about those
[6:37 PM] beta_helix:
https://fold.it/portal/node/2010706
Coronavirus Designable Linker Puzzles
Coronavirus Designable Linker Puzzles
[6:38 PM] formula350:For in-game clickers:
[6:38 PM] formula350:https://fold.it/portal/node/2010706
Coronavirus Designable Linker Puzzles
Coronavirus Designable Linker Puzzles
[6:38 PM] beta_helix: Thx
[6:38 PM] formula350:(once Josh's Quality of Life build gets debugged, that won't be a problem, heh)
[6:39 PM] donuts554:Also why is there a cluster of histidine amino acids to one side in the locked and unlocked part of the Aflatoxin Challenge puzzle?
[6:39 PM] beta_helix: @donuts554 I was just about to answer you
[6:39 PM] alcor29: Out of left field: Is anyone working on trying to do a Cas9 (CRSPR) attack on the cov-sars-2 DNA?
[6:39 PM] beta_helix: I'm looking at your screenshot…
[6:40 PM] jumper2: It would be nice to know how the DLP's react under specific circumstances. Looking at the screen where one's thumb is about an angstrom long, there is a mental model that is easy to fall back on of treating everything like levers and boards attached with string and normal non-quantum stuff. But would a DLP actually act like a monkey wrench thrown into the bearings of a steam turbine or would it act like a mote of dust. Good to understand the order of magnitude of these interactions
[6:43 PM] Susume: @alcor29 cov-sars-2 is an RNA virus (no DNA) - I think CRISPR for RNA is still very new.
[6:43 PM] beta_helix: @donuts554 I've been staring at that screenshot for a while and I don't see an crazy unusual amount of HIS there
[6:43 PM] jumper2: Another interesting application of DLP's would be if you could choose bonding sites that are conserved across mutations and have the middle stuff "over span" the area that sees all the mutation churn that confuses the immune system
[6:44 PM] beta_helix: (I counted 3 towards the bottom?)
[6:44 PM] alcor29:Thanks Susume.
[6:44 PM] beta_helix: That would be sweet though!
[6:47 PM] jumper2: Both measles and influenza are RNA viruses that mutate so fast they become almost a swarm of cousins. But the site the immune system binds to on measles doesn't change (or when it does the virus is not viable), but influenza just churns. But a DLP shaped like a horseshoe magnet could connect to points that don't change making some sort of super-stretch anti-bodies
[6:48 PM] donuts554:There are 6 histidines here
[6:48 PM] donuts554:IMAGE: http://fold.it/portal/files/chatimg/irc_902783_1604627324.png

[6:49 PM] beta_helix: Thanks donuts, I've been staring at zoom meetings all day. This helps!
[6:49 PM] formula350:That's not a very good screenshot to display them Donuts, since 2 of them are vertically viewed.
[6:50 PM] beta_helix: (although I keep trying to rotate your screenshot!)
[6:50 PM] formula350:We all do that Beta T_T Thus my request for our screenshots to be MOLs lol
[6:50 PM] beta_helix: Josh, you're totally on that, right?
[6:51 PM] sethcooper:there are also some updates to the blog post on metrics:
[6:51 PM] sethcooper:https://fold.it/portal/node/2010472
Introducing Foldit Metrics
Introducing Foldit Metrics
[6:52 PM] formula350:(apparently jflat has something in the works for us to view our proteins on the website using a MOL viewer, but the impression I got was it won't be screenshot related… time will tell!)
[6:52 PM] beta_helix: @Formula350 jflat totally does, but I think that won't be until after the new website is up and running
[6:53 PM] formula350: Yea :(
[6:55 PM] beta_helix: @Donuts My guess is that those are oxidized histidines, but I will check with UC Davis and post their answer in our post-Office Hour thread: https://fold.it/portal/node/2010531
[6:56 PM] formula350:https://fold.it/portal/node/2010531
[6:57 PM] donuts554:How do you think those clusters contribute to the aflatoxin enzyme's function?
[6:58 PM] beta_helix: are there any final questions for Seth or I? We will try to setup a Designable Linker Puzzles (DLP, copyright beta_helix: 2020) chat with neilpg628 very soon… probably after looking over the results of the first puzzle.
[6:58 PM] formula350:Something I've been trying to ask at Office Hours lately: What is one of the most common things you see us doing with our designs that either: you wish we would cut back on -or- you think is really interesting and would like it if more of us did it? (your choice; answering both is always welcomed lol)
[6:59 PM] beta_helix: @Donuts that would be my guess, but I want to get you a proper detailed answer from the experts… plus he was a homie of mine when we were both in the Baker Lab
[6:59 PM] beta_helix: @Formula350 Yes, I remember you always ask that, but totally I think I have a decent answer for you!
[7:01 PM] beta_helix: Now that we have the new metrics (that everyone loves ) in our mind those solutions are already useful. ie you don't need to keep grinding on them.
[7:01 PM] Skippysk8sIRC: you do know we play the last day for fun and group bragging rights. Is there anything more fundamental
[7:02 PM] beta_helix: Ideally, we'd love it if once you create a model that "satisfies all the metrics" you completely forget about that topology and come up with a brand new one from scratch!
[7:02 PM] formula350:In other words: "Don't End Game Your Designs Please"
[7:02 PM] beta_helix: @Skippysk8s I know but formula350 asked!
[7:02 PM] jumper2: produce a miracle, rinse, repeat
[7:02 PM] beta_helix: there you go! easy as that!
[7:03 PM] formula350:Mommy said I'm her little miracle ^_^
[7:03 PM] Skippysk8sIRC: ha ha. If there is feedback you can share, either positive or negative, we'd appreciate it
[7:03 PM] beta_helix: Why are we so picky? Because you all know the success rate in the lab for designs…
[7:03 PM] beta_helix: so even though your Foldit score might beat everyone else… it might totally crash out in lab
[7:04 PM] formula350:I think that's why we need more "Hand Fold" puzzles. I'm not sure why those went extinct :\
[7:04 PM] beta_helix: but if you generate 100 models that "meet the metric criteria" then that increases the odds A LOT
[7:04 PM] beta_helix: Seth, why did we stop those!?
[7:04 PM] jumper2: And it also seems like you want structure and repeatability and not a "just works by chance" like what evolution comes up with. Eventually, you want to bring out a protein kit. Hence the support for recipes
[7:04 PM] alcor29: So is anyone trying to figure out how to reward that behaviour: Giving up score to design another one?
[7:04 PM] Skippysk8sIRC: agreed. If possible, try to stagger end time for big puzzles so we can work on one by hand, put on recipes for game while we work on the other. I can't work 2 big puzzles at the same time
[7:04 PM] beta_helix: Especially for design, which I always saw as a lot more hands on (compared to running mutate forever)
[7:05 PM] jumper2: Because with recipes, we explain our logic for a repeatable process
[7:05 PM] beta_helix: @Skippysk8s very good point… we don't want to bring back any CASP memories (for those who survived those days)
[7:05 PM] Skippysk8sIRC: do you want us to share our "bomb" scores that might be ok
[7:05 PM] formula350:Ok this is an important followup to what you've said: If you'd like us to stop once we hit a good metric… What if we've done so under a Clash Importance of 0.40? In real life won't that want to relax more? So, shouldn't I work on it at a higher CI after, before submitting the design?
[7:06 PM] beta_helix: @Jumper2 that totally makes sense… but what if that logic has a flaw, and you might discover a better logic manually?
[7:06 PM] Dudit: Is it possible to combine DDG, SASA, SC into One Single Metric?
[7:06 PM] jumper2: New version of a recipe or a completely new recipe
[7:06 PM] beta_helix: @Skippysk8s honestly: if they don't meet the metrics, they almost have no chance of working in the wet lab
[7:07 PM] Skippysk8sIRC: no, I mean lousy score as it wasn't worked at higher wiggle
[7:07 PM] jumper2: The recipes are based on what we discover by manual folding
[7:07 PM] beta_helix: @Formula350 that is a good point… you could also make a garbage structure with decent metrics, and we don't want that. Again: we're only asking for perfection here, what's the big deal?
[7:08 PM] donuts554:Not all of the recipes
[7:08 PM] jumper2: Recipes are genetic plasmids in a sea of stinky bacteria all trying to get the best score
[7:08 PM] formula350:hahah
[7:08 PM] Enzyme2: Ihate endgaming more than anyone. But I like points more
[7:08 PM] robgee: points are my metric ;p
[7:08 PM] sethcooper:possibly metrics could be combined
[7:08 PM] beta_helix: @robgee and Foldit is a game, so it should be!
[7:08 PM] sethcooper:but is it more useful to see them separately?
[7:09 PM] alcor29: So the recommended metric scores are absolute minimum reqs?
[7:09 PM] beta_helix: We're honestly been discussing different ways of incentivizing what I've been talking about, because it's obviously not straightforward.
[7:09 PM] formula350:I guess my quesiton more was: is a fully-mutated and worked out solution at 0.40 CI still viable, or is wiggling at 1.0 a must?
[7:09 PM] Skippysk8sIRC: many of us try the science first. Then we play for fun, points and bragging rights
[7:09 PM] beta_helix: @alcor29 Yes, for sure… without any other problems with the model too!
[7:09 PM] jumper2: I like as much stuff separate as possible.
[7:10 PM] beta_helix: @Skippysk8s which is what the Share with Scientist button is for!
[7:10 PM] jumper2: More information display in densest form. Otherwise you might as well just have the top score and tell everyone if you want more detail that's what the "tab" key is for
[7:10 PM] Skippysk8sIRC: I do that… and label junk accordingly ;)
[7:10 PM] beta_helix: Nobody is going to stay with Foldit if you're not having fun… so we never want to take that away!
[7:10 PM] Susume: @Formula350 low CI lets things get closer together, so I'd be worried that the metrics might not hold when wiggling at 1.0
[7:11 PM] donuts554:How would the the Foldit score function look like if it was put in terms of the fundamental constants of this universe?
[7:11 PM] alcor29: How about the scores not postable if the metrics aren't met?
[7:11 PM] Skippysk8sIRC: hardball :)
[7:11 PM] beta_helix: @Jumper2 exactly…. we don't want to make the metrics a black box
[7:11 PM] formula350:We use it don't worry lol We've actually had to request the amount of solutiosn we can Share be INCREASED, since we've been making so many that we've ran out of slots lol (They have been increased though, so no worries. I think we're able to each upload 15 now, instead of 5)
[7:11 PM] sethcooper:I think I need to head out now
[7:12 PM] Skippysk8sIRC: don't do that to beginners
[7:12 PM] beta_helix: Thank you, Seth!
[7:12 PM] sethcooper:thanks everyone!
[7:12 PM] formula350:Thanks Seth, and Beta. Sorry for keeping you longer :}
[7:12 PM] Skippysk8sIRC: Thanks Seth
[7:12 PM] donuts554:You are welcome!
[7:12 PM] Skippysk8sIRC: Thanks Beta
[7:12 PM] Susume: hold up, Rosetta energy units aren't the fundamental constants of the universe? darn!
[7:12 PM] Dudit: How about gradual New Metrics points?
[7:12 PM] Enzyme2: if a player creates several design puzzle solutions. How many of those do you actually see?
[7:12 PM] formula350:@Dudit 1913 has that!
[7:13 PM] robgee: Thanks Seth and beta_helix:
[7:13 PM] jumper2: Thanks!
[7:13 PM] Skippysk8sIRC: Enzyme, I think we have to share one of each with scientists to be sure. It need not be end gamed all the way through
[7:13 PM] Dudit: Thank you
[7:13 PM] beta_helix: @Enzyme2 we see all the Share with Scientist solutions All of them! (even the junk-labelled ones)
[7:14 PM] donuts554:Yes but how would the equation to calculate the Foldit score function look like if it was put in terms of the fundamental constants of this universe?
[7:14 PM] formula350:Agreed Susume. I think that's what I was slightly curious to make sure of. Mainly trying to determine if designs I run out of time on, since I fold at 0.40, if I should bother to share it. By the sounds of it: No, it wouldn't be worth it, since it'll not be producible in real life.
[7:14 PM] Enzyme2: so if it's isn't shared, you'll only see the top solution. got it.
[7:15 PM] beta_helix: @Formula350 exactly… The wet lab is at CI = 1.0
[7:15 PM] jeff101: https://fold.it/portal/node/988112#comment-28798
[7:15 PM] beta_helix: @Enzyme2 in general, I think this is a fair statement… though that depends on the puzzle, of course
[7:15 PM] jeff101: maybe rmsd dot plots of each player's solutions would encourage more diversity of solutions
[7:16 PM] donuts554:Maybe there could be an all loops challenge where there cant be any sheets or helixes in the protein
[7:16 PM] beta_helix: Does anyone remember the Exploration Map?
[7:17 PM] Skippysk8sIRC: yes. It was easier than CASP
[7:17 PM] beta_helix: Hahahahahhahahahahaa
[7:17 PM] beta_helix: It started sucking up a lot of memory, though!
[7:17 PM] formula350:We have plenty of spare (computer) memory these days.
[7:18 PM] Skippysk8sIRC: so try totally different structures for you if possible
[7:18 PM] formula350:Bring back Exploration klthnx<3
[7:18 PM] alcor29: I still think the obvious is: If metrics aren't met the scores are not valid.
[7:19 PM] Susume: donuts our designs really have to have SS in order to fold up. Nature can get away with long loops but we can't
[7:19 PM] formula350:Honestly, I'm… onboard with that Alcor!
[7:19 PM] Skippysk8sIRC: need new players alcor29. Though I agree that after a while your suggestion is fair
[7:19 PM] beta_helix: Ok gang, I gotta go… (I just saw your "fundamental constants of this universe question", donuts! I don't know if I can ask around for any answer on that one… but I'll try)
[7:19 PM] Susume: thanks beta
[7:19 PM] alcor29: Yeah but those players usually give up anyway. Maybe/
[7:19 PM] formula350: Skippy, Metrics aren't a part of the BEginner puzzles, so that won't scare off newbies.
[7:20 PM] alcor29: Thanks beta
[7:20 PM] Dudit: Thanks @beta_helix
[7:20 PM] beta_helix: @alcor29 it might be a different type of puzzle, like a Boss Level or something.
[7:20 PM] Formula350: Thanks Beta. Try to get our Hand Fold puzzles back lol
[7:20 PM] alcor29: Hahahaha
[7:20 PM] jeff101: thanks beta
[7:20 PM] Skippysk8sIRC: Thanks Beta. We do need to let our new players get some reward or see some progress as they move to the regular puzzles
[7:20 PM] alcor29:Thought that might work. Having another review level.
[7:20 PM] beta_helix: I'll bring up hand puzzles for sure, I always loved those… maybe we can make them last 1 month, so there is no time pressure.
[7:21 PM] beta_helix: Thanks again everyone, keep up the great folding!

[agcohn821]

11/14/20:
pre-office hour
11:39 AM] : HuubR: Question for Neil: What is the most important aspect of the linker: just to keep the two binders at a fixed distance, or also to make sure their orientation stays the same (relative to each other). In other words, when the binders can pivot around the ends of the linker, is that good or bad?
[11:39 AM] jeff101: yes, thanks Ariana. Have a good weekend.
[11:40 AM] HuubR: Yes, Ari, thank you for spending time on a Saturday morning.
[11:46 AM] jeff101: I put zero length bands of strength 10 in the non-mutable part of 1918b to keep them in place.
[11:46 AM] ZeroLeak7: results of the last rounds of the MERS-CoV Binder Design in the lab?
[11:46 AM] jeff101: I use the reset puzzle pose as a Guide to make sure things don't move.
[11:47 AM] ZeroLeak7: yes but the clashes are there
[11:47 AM] jeff101: It makes building the linker more predictable, but maybe we are supposed to let things move more freely.
[11:48 AM] jeff101: if I omitted the zero length bands, would the clashes go away?
[11:48 AM] ZeroLeak7: no it's on the locked lcb1 binder
[11:50 AM] jeff101: so we are stuck with the clashes that are in the non-mutable parts, right?
[11:50 AM] ZeroLeak7: yes
[11:50 AM] ZeroLeak7:
[11:50 AM] donuts554: hello
[11:51 AM] jeff101: hi donuts
[11:51 AM] ZeroLeak7: hi
[11:51 AM] Todd6485577: Morning, well afternoon everyone.
[11:51 AM] ZeroLeak7: hello todd
[11:52 AM] donuts554: hello how are you?
[11:53 AM] ZeroLeak7: I'm fine thank you and how are you?(edited)
[11:53 AM] Todd6485577: Doing alright, working on A+ cert stuff and three puzzles.
[11:53 AM] Todd6485577: Hows everyone else doing?
[11:54 AM] donuts554: I am good, thanks for asking!
[11:54 AM] ZeroLeak7: yes I'm doing great on the all 3 puzzles
[11:54 AM] Todd6485577: also running powershell cipher command on my work computer to scrub all my deleted files
[11:55 AM] ZeroLeak7: are you in gargleblasters now :)?
[11:56 AM] Enzyme2: I think he just needs to restart Foldit
[11:56 AM] ZeroLeak7: ok
[11:56 AM] Todd6485577: Yeah, just joined it today. Needed to be able to see more activity to eveolve.
[11:56 AM] Enzyme2: shows him on the website, but not in yhe group chat
[11:57 AM] ZeroLeak7: gratulations
[11:58 AM] Todd6485577: I'm doing alright on the three puzzles, I'd like to see others solutions so I could be better…hence the group change. I'm pretty sure that I was almost the only one active in hold my beer
[11:58 AM] ZeroLeak7: I bet enzyme could show you some new things
[11:59 AM] Todd6485577: Enzyme, you, formula, I notice who's been active in vet chat
[11:59 AM] ZeroLeak7: there are great players with a great knowledge
[11:59 AM] donuts554: Why is there a Residue Limit on the number of residues in 1918b?
[12:00 PM] Enzyme2: All I do is shake and wiggle
[12:00 PM] Todd6485577: I mutate and wiggle
_____________________________________

11/15/20:

[10:59 AM] neilpg628: Hi all, its neilpg628 ready to answer any questions about Designable Linker Puzzles, and really any other questions about Foldit!
[11:02 AM] HuubR: Hi Neil
[11:02 AM] neilpg628: Hello!
[11:04 AM] HuubR: I certainly hope it's not just the two of us, but I'm glad you could make it today.
[11:05 AM] neilpg628: People should be dropping in soon Do you have any questions?
[11:05 AM] donuts554: hello
[11:05 AM] HuubR: And yesterday, it was good to hear that you were OK, it was just your car that let you down :-)
[11:05 AM] pc: In linker puzzles, to have something less flexible, is it interesting to maximise interactions (boud, hydrphobics) between the linker and the 2 binders in gray ? Like create a helix near one of the binder helix ?
[11:06 AM] donuts554: Why is there a residue limit on the linker?
[11:06 AM] pc: hi ^^
[11:06 AM] donuts554: hello how are you?
[11:07 AM] neilpg628: @pc Maximizing interactions between the linker and binders is definitely something to have, since that helps provide rigidity to the structure. There were some designs that had this from previous puzzles that we are computationally analyzing
[11:08 AM] neilpg628: The residue limit is primarily to prevent slowdown in Foldit. The current puzzle is one of the largest we've ever had (max length=250) and even though much of it is locked, we don't want to make the puzzle unplayable
[11:08 AM] pc: thanks. In 1918b there is some limitations. We can't create an helix here for exemple, because the 77 sidechain isn't changeable and make clash. http://fold.it/portal/files/chatimg/irc_959223_1606053501.png(edited)

[11:09 AM] pc: Maybe an extended helix here can be interesting to create something less flexible after.
[11:09 AM] pc: (clash, not crash ^^)
[11:10 AM] HuubR:https://fold.it/portal/files/chatimg/irc_959223_1606053501.png

[11:10 AM] HuubR:(for in-game chat)
[11:10 AM] neilpg628: @pc I think that is a valid concern. Originally we did leave the sidechains on the binders unlocked, but that caused slowdowns from the sheer numbers. Possibly we could let only some be flexible for this reason All of the frozen bits are directly from experiments done in the IPD, so those are real results
[11:11 AM] HuubR:So if I understand you correctly, the task of the linker is not just to keep the two binders at a fixed distance, but also to make sure their orientation stays the same (relative to each other). Is that correct?
[11:12 AM] donuts554: Oh yes the orientation too
[11:13 AM] neilpg628: Yes HuubR, that would be ideal. Both of the binders bind to the SARS-CoV-2 spike extremely well, but if we could link them rigidly, we would automatically have a much better binder(edited)
[11:13 AM] neilpg628: The way they are currently oriented matches how they would be in real life
[11:14 AM] HuubR:That's a clear answer. But I have been wondering if there is any way in which we can assess how rigid a particular design is?
[11:16 AM] neilpg628: That would be done on our end, there is no one metric for 'rigid' However, things with good secondary structure, packing, ideal loops, etc (In general things that give a good score) are better contenders
[11:17 AM] HuubR:Can you give any feedback on what we produced in the first rounds, or is that too early?
[11:18 AM] HuubR:(I know that this is a new type of puzzle for all of us :-)
[11:18 AM] donuts554: Should the linker protein be mostly in the concave side of the two locked parts, or be mostly in the convex side of the two locked parts?
[11:20 AM] neilpg628: The convex side seems more likely, since you have more space, but honestly its up to you! There simply is not much data on these types of design problems
[11:22 AM] neilpg628: Also HuubR, we have looked quite a bit at the first two, and I would say that the best designs tend to have a lot of interaction between the linker and the binders (since this helps maintain a fixed structure for the linker) Some of the top structures have a-helices on the linker that form hbonds with the residues on the binder
[11:22 AM] pc: Is it a problem if our linker make some bonds with the virus ?
[11:24 AM] neilpg628: Probably not, though it might be hard to say without more rigorous testing. We would want the linker to behave the same regardless of the virus presence though, because the LCB1-linker-LCB3 complex should act as an independent unit
[11:24 AM] pc: thanks. Are hydropholic bouds in our linker are strong enough to have something less flexible, or hydrophobics interactions are more interesting for that ?
[11:26 AM] neilpg628: @pc I'm not sure if I follow? Generally we would expect that the core remains mostly hydrophobic
[11:27 AM] susume: would it possible to place the boundary between locked and movable residues at a pair with good helix dihedrals? - then we could continue the helix or switch to loop
[11:28 AM] nspc:IMAGE: http://fold.it/portal/files/chatimg/irc_959223_1606069682.png

[11:28 AM] neilpg628: @Susume We actually discussed this on Thursday, and will be doing this for the next puzzle!
[11:28 AM] susume: cool, thanks!
[11:28 AM] nspc:(I am pc) I mean this kind of bouds maybe are too much flexsible and we have to make more hydrophobics interactions instead
[11:29 AM] neilpg628: I think that looks fine as is
[11:31 AM] pc: thanks, so hydropholic networks can be enough to have something strong in some parts (and not too much flexible)
[11:32 AM] donuts554: But I think the hydrophillics want to be around water
[11:32 AM] pc: so it can be less stable than hydrophobic if they want to do boud with water instead ?(edited)
[11:34 AM] neilpg628: @pc Hydrophobics are good for the core, since they allow for a core that actually folds (since hydrophobics want to interact with eachother more than water) Hydrophilics would be good for making hbonds with other residues and water
[11:39 AM] pc: ok thanks. So hydrophobics are mostly here to help the folding. And hydropholics with bonds keep the structure strong.(edited)
[11:40 AM] pc: Is a too long helix is too much flexible ?
[11:41 AM] donuts554: Can the two locked binder parts be independent at first, and then can they rigidly bind to each other and form a disulfide covalent bond on the inside center of the binding site which cant be destroyed, so it still acts like a whole protein at the end?
[11:44 AM] neilpg628: A long helix is not necessarily a bad thing, as long as it is making proper interactions
[11:45 AM] pc: is the main idea of the linker puzzles is to increase contact surface (like sasa) by having 2 binders ? And make it more sticky ? Or there is for other things, like have more chance to find the binding zone ?(edited)
[11:45 AM] pc: thanks
[11:46 AM] Formula350: Just sat down (though did read everything). My quick question before the buzzer: Just curious if the LCB1+Linker+LCB3 will be expressed as a single protein, or if they are intended to be expressed individually and expected to bond as we've designed?
[11:46 AM] neilpg628: donuts554, I'm not sure if there are any disulfides to be formed. If there are no cystines near the binder surface I doubt we could make them, since those binder AA sequences are fixed
[11:47 AM] neilpg628: @Formula350 They would be expressed as a single protein since they are a single AA sequence. We wouldn't be attempting to splice sequences together
[11:48 AM] neilpg628: @pc The main idea is to make a better binder by joining two others. If we could lock the two binders in the correct orientation/distance before binding the resulting protein would have a much higher affinity because the surface area to contact would be much greater
[11:49 AM] Formula350: Thanks. I didn't think so, but I also thought that Aflatoxin puzzle that we were working on was a single chain, but apparently that's a 'ligand'. So finding that out kind of changed my impression of how things could actually be used in the lab vs what we're seeing in Foldit
[11:50 AM] Formula350: Speaking of what pc said, would the CSM Objective be something applicable to the Linker puzzle?
[11:52 AM] pc: @Formula350 for me CSM is only for the part that is in contact with virus, so not very usefull for linker I think. (But it can be usefull if we add part that touch virus maybe..)
[11:53 AM] pc: but, maybe it can be better than void and packing
[11:53 AM] Formula350: Well yes, I suppose what I mean is a modified CSM that was meant to look at only itself.
[11:53 AM] pc: for the linker
[11:53 AM] neilpg628: Do you mean CMS?
[11:53 AM] neilpg628: Contact Molecular Surface
[11:53 AM] Formula350: Contact Surface Metric, the new combo SASA/SC in 1920
[11:55 AM] neilpg628: Oh yeah, CMS is the scientific name That could be useful but we would only add that later as there are still some kinks to work out
[11:55 AM] neilpg628: The Aflatoxin Puzzle does have a ligand, but in this case everything is protein, so that is not an issue
[11:55 AM] HuubR:So if I understand you correctly, the task of the linker is not just to keep the two binders at a fixed distance, but also to make sure their orientation stays the same (relative to each other). Is that correct?
[11:56 AM] HuubR:(Sorry, wrong clipboard)
[11:56 AM] HuubR:I've been looking at the puzzle with what you just told us, about LCB1+linker+LCB3 being a single protein. I would say that the combination is far from being a "globular" shape. Is there a risk that the total protein will not fold like we see it in the puzzle, but rather into something more compact?
[11:56 AM] Formula350: Ah ok, good to know, sorry for the confusion then lol I enjoy the new metrics, as for me, they've been very helpful to create closer to what you all want. So I look forward to being able to use them more (Though I know not everyone does, since older computer are having a tougher time with them)
[11:57 AM] HuubR::D
[11:57 AM] pc: neilpg628, The final protein of linker puzzles look likes an Antibody. Is there some ideas to make something like an antibody later (to be compatible with "phagocyte cells" for exemple)
[11:57 AM] Formula350: I share HuubR's concern and curiosity there, too, which was also partially why I thought perhaps they would be expressed individually, to prohibit them folding properly.
[11:58 AM] Formula350: Err wording. prohibit folding improperly.
[11:58 AM] donuts554: I mean the binder proteins and the linker, not just the binder proteins. I mean can the linker be cut in half, and so one half of the linker tries to rigidly bind with the other half of the linker, and so that a disulfide covalent bond is formed in the inside center of the binding interface of the two linkers, which cant break, so that the two parts could act as one independent protein in end?
[11:59 AM] neilpg628: HuubR, We won't know for sure what anything looks like folded until we make these sequences, but since the binders fold independently into their proper shapes it is unlikely that the whole thing would form a globular cluster
[12:00 PM] neilpg628: @pc These probably wont be used as an antibody, since those are only produced by your immune system
[12:00 PM] HuubR:You mean the binders have strong enough interaction within them to fold as sub-units, and then the whole protein should fold as we see it in the puzzle?
[12:01 PM] neilpg628: donuts554, I can ask about that, it sounds interesting
[12:01 PM] neilpg628: HuubR, That is correct, and we know from experiments that the binders fold as intendend
[12:01 PM] HuubR:Thanks, good to know that!
[12:04 PM] donuts554: Cant there be a dimer puzzle, and the highest working dimers have each protein attached to each of the binders, so they can bind in the end and act as a single protein?
[12:04 PM] Formula350: Yea I think what Donuts is after is closer to the initial test puzzle for linker that we had gotten a number of weeks back, which had twin locked helices, and we had to try to come up with a way to bond them. In that one, the two bundles (and thereby, what we could manipulate) were separate. So my assumption here is that he's hoping we could use CYS to bind and strengthen our linker, by having it be LCB1+half the Linker, and then the other half of the linker+LCB3. (sound about right, Donuts?)
[12:06 PM] donuts554: Yes it sounds about right
[12:06 PM] susume: donuts, disulfides are strong, but don't forget the peptide bonds (the ones that hold two pieces of backbone together) are super strong too, so having it all be a single protein is quite strong - we just have to get a stable shape for it
[12:07 PM] susume: heh, 'just'
[12:07 PM] neilpg628: I can ask about that, but as Susume said, the actual covalent bond between bb residue atoms is as strong as any disulfide
[12:08 PM] Formula350 Foldit Players - We make the Impossible….. a bit more Plausible!
[12:08 PM] susume: +1 plausibility
[12:09 PM] neilpg628: Exactly
[12:09 PM] donuts554: Oh ok, I thought that the linker problem could be solved by attaching a half of a hemoglobin complex/DNA to one binder side and the other half to the other binder side
[12:11 PM] neilpg628: That could get complicated, since we also have restrictions on avoiding the virus
[12:11 PM] neilpg628: Anyway, I think we are good here!
[12:11 PM] neilpg628: It was nice talking to you all!

[agcohn821]

[2:22 PM] jflat06: Hey all, I'm here now
[2:22 PM] dudit:: Hi @jflat06
[2:22 PM] jeff101: Yay!
[2:22 PM] formula350: Oh so maybe he was applying Waffle Face like I was going to joke about :D
[2:22 PM] formula350: (sleeping on his keyboard)
[2:23 PM] jeff101: Did you attend the CASP meeting this past week?
[2:23 PM] jflat06: (was in the middle of making grilled cheese)
[2:23 PM] jflat06: I did not
[2:23 PM] jflat06: We talked about it at our meetings, though
[2:24 PM] grogar7: Alphafold sounds like a huge breakthrough
[2:24 PM] formula350: Technically "Alpha Fold 2"
[2:24 PM] dudit: What is the future of Foldit when AlphaFold is capable in Protein Design Puzzle?
[2:24 PM] pc: jflat06 : are LCB1 and LCB3 usefull alone for a medication against covid ? And when they can be usable on humans ?(edited)
[2:25 PM] jflat06: Indeed it does. As a computer scientist (instead of a proper biochemist), I had to clarify with some of the other foldit team as to exactly how much of it was hype and how much was not.
[2:26 PM] spvincent: Amazing achievement though AlphaFold2 is, does it provide any extra insight or understanding of the protein folding process?
[2:26 PM] jflat06: @pc i'm not up to date on the status of the various antivirals
[2:27 PM] jflat06: Probably a better question for a biochemist - my guess is it would still be a while, unless they're already undergoing human trials for safety
[2:27 PM] jeff101: what have you been working on lately that you can tell us about?
[2:28 PM] formula350: Yea, In other words, the best Qs to ask jflat would revolve around the software and tools of Foldit or similar. (As I understand it heh)
[2:28 PM] jflat06: @Dudit: We haven't seen exactly what alphafoldit will look like in protein design, but we're starting to think about ways to incorporate it in a way that still leverages humans
[2:28 PM] jflat06: haha, alphafoldit (i can't type fold without the it, apparently)
[2:28 PM] grogar7: lol
[2:29 PM] formula350: Freudian Slip, perhaps… Might be subconsciously answering Jeff's Q…. ;) ;) lol
[2:29 PM] formula350: err jeff101's, since, you are also a Jeff…
[2:30 PM] jeff101: or what do you plan to work on next week?
[2:30 PM] dudit: @jflat06 so Foldit will still exist forever?
[2:30 PM] susume: to what extent has the team talked about adding trRosetta info to foldit? (trRosetta is based on alphafold 1)
[2:30 PM] Josh:
Ooh! susume: I know this one! But I can let jflat answer :)
[2:30 PM] jflat06:
If alphafold really turns out to be an oracle for protein design, the hardest part of folding will shift to accurately describing the problem at hand, so that the neural nets can work their magic on that.(edited)
[2:31 PM] Jumper2:
Someone mentioned Alphafold on Github, who's up for writing the alphafold.lua recipe? LOL
[2:31 PM] jflat06:
So one possibility for the future is to have foldit players design puzzles for a future design-alphafold to work on
[2:32 PM]
formula350: I could also see us still being allowed to then attempt to improve upon what the AF2 designs, to further its own 'knowledge'.
[2:32 PM] pc: jflat06 : Is alphafold team is interested by foldit player solutions database ? (with all solutions we do)(edited)
[2:32 PM] susume: we could design backbones with starting sequences, and alpafold-design could find good sequences for them - because backbone variety and novelty is a foldit strength
[2:32 PM] jeff101: does alphafold work for man-made sequences?
[2:33 PM] jflat06: I think alphafold mostly uses native backbones as their sources for training data
[2:33 PM] jflat06: @Susume We're definitely pushing for trRosetta into Foldit - it's an active project!
[2:33 PM] spvincent: Yes, wouldn't Foldit designs likely have no analogues in AlphaFold2s training set?
[2:33 PM] susume:: jeff101 I want to know that too - alphafold 1 worked well on designs, can't wait to hear if alphafold 2 does too
[2:34 PM] jflat06: @spvincent depends on the particular design, but yeah, a lot of foldit designs are fairly novel
[2:34 PM] jflat06: the cool part is if we verify your structure experimentally, it IS a native structure at that point, and could be included in alphafold's model
[2:35 PM] susume:
https://fold.it/portal/blog?page=1
Newsletter July 3: Initial Reactions
Newsletter July 3: Initial Reactions
[2:35 PM] Jumper2: If both are neural net/AI type efforts, is there a succinct comparison of Alphafold to trRosetta?
[2:35 PM] spvincent: ok, tx
[2:35 PM] susume:reminder of the trRosetta paper showing it works well on design
[2:36 PM] jflat06: I believe trRosetta is based on the original alphafold from a while ago
[2:36 PM] jflat06: the new methods in alphafold2 will presumably be included in an update to trRosetta
[2:36 PM] Jumper2: trRosetta++
[2:36 PM] jflat06: hah
[2:36 PM] Jumper2: with templates
[2:37 PM] jflat06: Anyway - to summarize, we are definitely thinking about neural nets, and how to incorporate them into Foldit
[2:38 PM] jflat06: science/development can be a bit slow, but you should expect to see something in the future
[2:39 PM] jflat06: As for my current work, I am working on getting the client co-operating with the new website. There's a bunch of changes to our server communication protocols that effectively require a re-write of all of our server code. And our server code isn't exactly… modular. It a big old messy hairball 13 years in the making.
[2:40 PM] Jumper2: One idea would be to give Alphafold and the Foldit players the same initial puzzle to see what the solutions are. If the players solutions seem better, then test in the lab. With lab results turn the player solutions into training inputs (both positive for working solutions and negative for those that fail to fold or end up in clumps). Then the foldit puzzles really become meta-puzzles of generating training scenarios.
[2:41 PM] Jumper2: That type of collaboration would seem like it would generate some interesting papers relevant to a wide swath of sciences
[2:42 PM] jflat06: Yeah, there's definitely the potential for producing some training data for a model.
[2:42 PM] jflat06: That's a bit hands-off, though, and it would be far more interesting to have players interacting with design even after AlphaFold has been properly trained
[2:43 PM] formula350: I have 3 questions, and the first 2 are rhetorical: 1) Where's Waldo? 2) Where in the World is Carmen San Diego? 3) If you think of the players like your kids, what is your "proudest dad moment" in your time while working on Foldit?
[2:43 PM] jeff101: do you ever code Foldit LUA commands?
[2:45 PM] jeff101: I'd like to see more related to the Objectives, like list the abego color for each segment, list which segments are in non-ideal loops, have the ideal loops filter depend on the # of residues in non-ideal loops instead of the # of non-ideal loops
[2:45 PM] jflat06: @Formula350 probably when our players discover or create something that makes even our scientists do a double take.
[2:45 PM] jflat06: @jeff101 yup! I think Josh/milkshake have been doing most of the recent work, though.
[2:46 PM] formula350: Doesn't have to be now, but if you happen to have any examples of that, I'd certainly love to hear about some! :D
[2:46 PM] Jumper2: The interesting question is one of what produces good training data. The idea that people and neural nets build up intuition based on repeated exposure to massive amounts of raw experience data. Would training data based on human player intuition be doubly useful for training or would it just end up confusing the AI because of seeming to be too random?
[2:46 PM] dudit: I think it will be great if every Foldit puzzle have a video tutorial in one place
[2:48 PM] jflat06: I do believe we have more to come on at least the objectives. The metrics are very new, and we're trying to make sure we have the right idea before going and giving you lua functions so that we don't have to change them in a way that would break your scripts later on. This is one of the biggest hurdles to writing lua functions.
[2:48 PM] jflat06: @Formula I would definitely have to think about it - I have a terrible memory, so it'd be hard on the spot.
[2:49 PM] formula350: Yep I can relate, if you think about it, I'm always on the Discord :P If not, no biggie!
[2:49 PM] jflat06: @Jumper2 That would be in interesting question in its own right.
[2:50 PM] formula350: That may 'taint' the results dudit:? I think they try to give us any specific "Do this and that" information, beyond what they do now, because they want us to come up with the things they wouldn't have. But I could be wrong.
[2:51 PM] jflat06: Yeah, we definitely like to try and influence players as little as possible going into a puzzle, as we really need as much diversity in the solutions as we can get
[2:54 PM] spvincent: Do you have any particular thoughts on triple helices in that regard?
[2:56 PM] jflat06: (Again, I'm not a biochemist!) I think for our freestyle design puzzles, those aren't always the most interesting. My understanding is that Rosetta does helical bundles well, and foldit players have 'conquered' that type of fold a while ago. However, I believe that in our targeted design puzzles (such as binders for coronavirus), we don't really care if it's some never-seen-before type of foldit - we just care if it works!
[2:58 PM] Jumper2: So what part of Foldit are you spending the most time on as a developer lately?
[2:58 PM] jeff101: more ideas for LUA functions: list blue green or orange for each segment to match the colors you see when you show the Core Existence Objective
[2:59 PM] jeff101: do you know if the Interaction Energy Objective is a good metric for pi-stacking?
[2:59 PM] jflat06: In a way, the freestyle design puzzles (and the nature paper they created) were where we refined foldit and our puzzles so that players could design all sorts of backbones. And now we're starting to try and apply that towards some real world problems.
[3:00 PM] jflat06: @Jumper2 I am working on the new website and associated client updates, as well as general foldit maintenance and tasks.
[3:01 PM] jflat06: @jeff101 I do not. I'm not educated enough on that to be able to say.
[3:01 PM] jeff101: will all the old content be preserved in the new website? like images we have put on the site?
[3:02 PM] jflat06: Most content should be preserved. I can't promise everything, but we're going to try to make it work.
[3:02 PM] formula350: (I'm almost positive this was mentioned in public, which has caused me to waffle with asking but…) The website feature that'd been mentioned once, where we'll be able to use a MOL viewer on the website to look at each others designs… Will that also extend to our shared Screenshots? Or will those still remain as only PNGs?
[3:03 PM] jeff101: will there be a transition period in which both the old and new websites will be online, but the old one will be read only?
[3:04 PM] jeff101: I can't imagine it all working right the first day. What precautions will you take so we can keep playing Foldit and using its website during the transition period?
[3:05 PM] jflat06: @Formula350 We aren't certain on all of the details there at the moment. We would like sharing to be as easy as possible, so you could imagine a 'screenshot' button that would post a link to the model on the website. However, that effectively means that other people could see your raw PDB, atom coordinates and all. So there are some concerns about exactly how this would affect the competition, among other things.
[3:06 PM] jeff101: I sometimes wonder if posting images of our solutions on the Foldit wiki is risky in the same way
[3:07 PM] jflat06: @jeff101 We plan to have a transition time where the old site will be read-only. There will also be a public demo of the website prior to we make the hard-switch.
[3:08 PM] formula350: Yea I had that same exact concern. But then I considered that maybe you guys would be able to use the encrypted file saves you are already, with a modified MOL viewer that'd be able to read them. Then everything is kept 'secret' still. As I'd seen on… one of the sites (YangLab's trRosetta server maybe, or perhaps ZhangLab's CoV-19 info DB) where their Viewer didn't permit the user to view that.
[3:08 PM] MikeCassidytoo: There could br three levels of competitiom: solo, evolver, shared evolver (or some other name)
[3:08 PM] jflat06: Unfortunately using the encrypted solution files isn't really possible.
[3:09 PM] formula350: Locked out most of the functions and even seemed to hide the file's URL behind an obfuscated link (as I'd viewed the page info)
[3:09 PM] formula350: Ah, shame.
[3:09 PM] jeff101: more LUA tools for Objectives: list which reisudes are scoring poorly for the Interaction Energy, and list which ones are doing poorly on the SS Design Objective.
[3:09 PM] jflat06: Whatever the website does to show them to you, they have to end up as a raw PDB at some point, so it would have to decrypt them in javascript right out in the open.
[3:10 PM] formula350: What about using WebGL or CL instead of JS? (which seems to run smoother than JS anyhow)
[3:10 PM] jflat06: webgl is a javascript API
[3:11 PM] jflat06: so your code all is still just code running on the end-user's machine.
[3:12 PM] formula350: hah Did not know that about WebGL! But yea, like I knew we're the side doing the "work" and need the file, I just hoped maybe it'd get stored in memory and need a ton of leg-work to extract.
[3:13 PM] Jumper2: Re: 3d viewer for screenshots Since it's already setup to hide the atom coordinates, it follows that the only viewer that should be able to parse and display the screenshots in 3d would be the foldit game client itself. As mentioned any other viewer will require the unencrypted data in order to render it and you guys have better things to do than write a separate 3d web viewer app on top of everything else. So if you limit 3d screenshots to being viewed in the foldit game client only, then you "simply" (hah, as if writing that makes it so) just add a "read/view only" flag to a game save file. Then 3d screenshots become much like saves. There could be links to them on the website, but the viewing would have to be in the game client. (Similar to recipes listed on the web but loaded into the client)
[3:13 PM] formula350: As, if that's the case, I personally wouldn't care if anyone went through that trouble to copy what I'd done lol (Though a major lab might be another story, however, I thought all of Foldit work taht gets published was basically freely available).
[3:13 PM] jflat06: Possibly. My favored approach would be to try and think about how to design a fair foldit competition framework that wouldn't be harmed by this sort of thing. That way we could really embrace sharing.
[3:14 PM] jeff101: sometimes I try new things on expired puzzles. if I am logged into Foldit, do solutions I generate then still get sent regularly to the server?
[3:14 PM] jflat06: i believe they do
[3:14 PM] jflat06: but it would be rare for us to look at them
[3:15 PM] formula350: Oh, or as Jumper said, bypass the browser and have the Foldit client "view" the solution! Which actually…
[3:15 PM] Jumper2: If 3d screenshots are developed as saves marked read/view only, then they could host other security flags such as limiting the viewer to being within a certain group
[3:15 PM] formula350: Having that capability would go very hand in hand with a feature request which would give us the ability to view our own Saves that way…
[3:15 PM] jflat06: As far as a web viewer goes - there's actually some already written that we have integrated already. So we wouldn't have to write that from scratch.
[3:16 PM] Jumper2: And because they would be read/view only, they shouldn't generate change deltas flowing back to the server or allow people to build their solutions directly on top of that file.
[3:17 PM] jeff101: what if within the Foldit client, there was a but like the Screenshot Camera icon that would rotate the protein around some axis with steps of like 10 degrees and make a gif file movie of the rotating protein?
[3:17 PM] formula350: Yea my thinking is to use whatever it is that the little preview window for Building Blocks has, just in a bigger window like Rama Map.
[3:17 PM] formula350: You probably got that from my suggestion on the Document, jeff.
[3:18 PM] formula350: Sug 1 is a MOL viewer, and falback Sug 2 was a low-FPS, low res GIF (slideshow almost, to keep file size down)
[3:19 PM] jflat06: In any case, we do plan on having a web-viewer for the new website. This is actually a huge request internally among our scientists, as this sort of thing would be invaluable for talking about puzzle results and clearly communicating structures and their strengths/weaknesses to players.
[3:19 PM] jeff101: would posting gif movies on the website be a more secure way to show the 3D structure of our proteins
[3:19 PM] Jumper2: One question would also be whether or not the read only flag would allow recipes to work at all (you could argue that the GetDistance LUA function called on a read-only 3d screenshot save would not be kosher)
[3:20 PM] Jumper2: Now when talking about scientists viewing it, I'm imagining them wanting to look at angstrom level measurements between residues and all the other things for which a full PDB file is needed.
[3:20 PM] formula350: If they open them in a "Building Block Preview" sort of window, which would be overlayed on our screen (sans transparency), there'd be literally no concern about that, since Recipes just flat out COULDN'T work.
[3:24 PM] Jumper2: When considering the detail that scientists might want to see combined with the ease of hitting "share with scientists". That argues for including a very simple way of translating a save to a PDB but that is probably restricted to admin usage only. That would allow "share with scientists" results to automatically be converted to a valid PDB format allowing use of a JavaScript or PyMOL type viewer. They would just have to be cautioned about not saving it in an area reachable by players.
[3:25 PM] jflat06: right, that's roughly our plan, at least for the initial rollout. We have talked internally a lot about how much of these capabilities we want players themselves to have, and we aren't necessarily opposed to the idea of letting players export their PDBs themselves.
[3:26 PM] jflat06: The key is that, if you can export a PDB, we need to ensure that importing one isn't easy.
[3:27 PM] jflat06: but as a first step, this solution-on-the-website thing would be an admin-only tool until.
[3:27 PM] Jumper2: Or if you can export a PDB you should ensure that such a function cannot be called on just any particular save. Thus read-only 3d screenshots would not be exportable, while full player saves might be
[3:28 PM] Formula350: lol Speedy "Proof of concept" (Kept purely internal to Foldit, no web use)

[3:28 PM] Formula350: Whoops Wrong screenshot
[3:29 PM] Jumper2: Import needs to be restricted regardless because it opens up the possibility of being able to use any PDB as a person's own solution.
[3:29 PM] jflat06: yeah, definitely
[3:29 PM] Formula350:

[3:29 PM] Formula350: Thhhat's what it was intended to be lol Dang Photoshop layers
[3:32 PM] Jumper2: But I definitely agree with allowing scientists easy access to player saves in PDB format. As players, IMO, I think anytime we hit the "share with scientists" button, we're in favor of anything that makes it easier for them to evaluate what we are sending them. That would include a lot of the cool features in PyMOL and other viewers they may be familiar with. I definitely would be against hamstringing them to subpar tools.
[3:33 PM] jflat06: just to clarify - all your solutions end up as PDBs in the end. It's just that the current pathway through which that happens doesn't make it easy to dynamically and quickly share these things. We definitely use PyMOL internally!
[3:35 PM] Jumper2: Perhaps in a future version when a client "shares with scientists" the server code automatically translates the save to standard PDB and saves it in a network share that scientists can easily get to with PyMOL or whatever tool they use.(edited)
[3:36 PM] jflat06: that's pretty much exactly what we are doing, lol.
[3:36 PM] Jumper2: As a teacher I once had always quipped "Great minds think alike, and fools never differ."
[3:38 PM] jflat06: haha
[3:38 PM] jeff101: I think the site the FOldit wiki is on will get updated too. Perhaps its address will change.
[3:39 PM] jeff101: When you update the Foldit site, will all the old links still work? Like if you have a bunch of Favorites in your web browser, will their addresses still work?
[3:39 PM] jflat06: the foldit wiki isn't hosted by us, so it shouldn't change
[3:39 PM] jeff101: I think they sent me an e-mail about upcoming changes.
[3:39 PM] jflat06: I think we're trying to preserve old URLs
[3:39 PM] jflat06: ah, ok
[3:41 PM] jeff101: I need to go do other things. Thanks for having this Office Hour.
[3:41 PM] jflat06: Yup, I think I am going to take off too. Thanks for all the questions!

[agcohn821]

[2:30 PM] milkshakeiii: it's officially office hours time 8)
[2:30 PM] argyrw: ok I left you
[2:30 PM] Dudit: Hi @milkshakeiii
[2:30 PM] argyrw: I hope to see photo of your design
[2:30 PM] Ridick051: yes :) welcome milkshakeiii
[2:31 PM] robgee: Hi milkshakeii, what the topic today ?
[2:31 PM] Formula350: Hallo. Glad you're OK :P
[2:31 PM] milkshakeiii: i can talk about the technical side of foldit
[2:31 PM] ZeroLeak7: can we ask questions?
[2:31 PM] milkshakeiii: yes!
[2:31 PM] ZeroLeak7: ok
[2:32 PM] ZeroLeak7: could you tell us more about what you are currently working on, is it just bug fixing or new features? and the second question is, if I copy something from the foldit editor and paste it into notepad ++, the formatting is wrong and only a long line of text is displayed. Will there be a fix for this anytime soon?
[2:33 PM] Formula350: Q2: If you're running DevPrev, then Josh will have to look into that again, but as far as I know it HAS been fixed already.
[2:33 PM] robgee: ZeroLeak7 you can use edit-EOL conversion in notepadd++
[2:33 PM] milkshakeiii: we don't like to talk about stuff that's in the early stages because we don't know exactly what form it will take by the time it gets to the players
[2:33 PM] milkshakeiii: today is my first day back from holiday and I'm working on a performance issue that we're seeing since around the time of the QoL patch
[2:34 PM] milkshakeiii: the second question is something Josh has been working on
[2:34 PM] milkshakeiii: I assume it has to do with line endings so you could try changing the line ending character in notepad++
[2:34 PM] Formula350: Yea it's fine in DevPrev, however, there are additional line-breaks included, but that's not so bad.
[2:35 PM] ZeroLeak7: ok ty yes it has to do with the line ending
[2:36 PM] Formula350: Editor to Notepad copy-paste results in DevPrev.
[2:37 PM] ZeroLeak7: oh good it has been fixed in DevPrev.
[2:37 PM] Dudit: @milkshakeiii Is there any performance difference between 32 bit vs 64 bit Foldit?
[2:38 PM] milkshakeiii:
Dudit 32 bit applications can only utilize a limited amount of RAM
[2:38 PM] milkshakeiii: However I believe Foldit should always stay under the limit
[2:39 PM] milkshakeiii: In general I would say avoid 32 bit applications
[2:39 PM] milkshakeiii: But in theory the performance shouldn't be too different
[2:39 PM] Huubr: Is Foldit on Windows a 64 bit application?
[2:39 PM] Formula350: Avoid… IF you have the option. lol
[2:39 PM] ZeroLeak7: no
[2:39 PM] Formula350: Cuz… we no haz the option for Foldit :P 32bit only on Windows.
[2:40 PM] milkshakeiii: i didn't actually know we were still posting 32 bit versions yeah
[2:40 PM] Huubr: Mine is in C:\Program Files (x86)\Foldit
[2:40 PM] milkshakeiii: i don't think we are (edit: nevermind)(edited)
[2:40 PM] Formula350: I know Mac's are 64bit (I think Linux as well), but Windows is still 32bit
[2:40 PM] Formula350: "20201223-e20c1e0f15-win_x86-devprev"
[2:41 PM] milkshakeiii: jflat would be the one who would know that
[2:42 PM] Formula350: What's your personal greated accomplishment with Foldit this year?
[2:42 PM] Formula350: greatest* (I can words)
[2:43 PM] milkshakeiii: probably implementing the metrics although they haven't been as useful as we'd hoped, but there's more possibilities for using that functionality in the future
[2:43 PM] ZeroLeak7: what do you think, you can get more performance out of foldit?
[2:43 PM] Formula350: "haven't been", in terms of producing better science results?
[2:43 PM] milkshakeiii: yes I definitely think I can get a lot more performance out of foldit
[2:44 PM] Dudit: @milkshakeiii Is Windows Foldit 64 bit version under development?
[2:44 PM] milkshakeiii: there's a lot of subobtimal behavior
[2:44 PM] ZeroLeak7: yeah awesome
[2:44 PM] Formula350: (Cuz persoanlly, aside from the performance, I love the Metrics.)
[2:44 PM] milkshakeiii: it terms of the asynchronous nature of them paying dividends
[2:45 PM] Trigger:The metrics need to be weighted properly. In one puzzle they were worth thousands of points, in others, a hundred. How important are they, compared to energy?
[2:45 PM] milkshakeiii: I'll have to ask about windows 64 bit, it feels a bit silly to be still making 32 bit versions in 2021, but it's a project with a long legacy
[2:46 PM] milkshakeiii: we want people to be highly motivated to reach a certain threshold on the metrics
[2:46 PM] milkshakeiii: but we don't want them to spend a lot of time optizing the value after it passes a threshold
[2:46 PM] ZeroLeak7: yes would be nice if on all platforms is 64 bit
[2:46 PM] Huubr: Did both of you mean "more performance" in terms of speed, or better results for science? The ideal situation would be to have both, of course, but if given the choice, I would go for science, not speed.
[2:46 PM] Formula350: LOL That's the nicest way I've heard the code be referred to… "a long legacy"
[2:47 PM] milkshakeiii: i agree science is better than speed, but hopefully speed can lead to better science
[2:47 PM] milkshakeiii: as long as it has all the same functionality just better ha
[2:47 PM] ZeroLeak7: yes that's it
[2:48 PM] Formula350: Yes, if you can do more in the same time, you'll tackle more possibilities, which can yield better science. And indeed, if corners aren't being cut, just optimization of code.
[2:48 PM] Huubr: I will probably need to speed up my brain, then. :-)
[2:48 PM] milkshakeiii: haha
[2:49 PM] ZeroLeak7: yes you can implant a neuro chip from elon musk
[2:52 PM] robgee: Does the foldit codebase keep up with modern algorithms ?
[2:53 PM] milkshakeiii: yes our protein folding algorithms come from rosetta which is under constant development
[2:53 PM] ZeroLeak7: ah in which version is it written of c++ ?
[2:54 PM] Trigger:Are any of the AlphaFold algorithms public domain, and if so, when will we see them added to the functions we normally use?
[2:54 PM] milkshakeiii: c++11
[2:54 PM] ZeroLeak7: thats old
[2:54 PM] milkshakeiii: not as old as Foldit!
[2:54 PM] Formula350: Would it be possible to port the Metrics calculations to OpenCL, to let the GPU calculate those instead of the CPU? And as a followup to Rob's… there was mention that Foldit would be getting a Rosetta update that would be talked about, but I haven't seen anything… That still happening?
[2:55 PM] milkshakeiii: the alphafold algorithms aren't public domain yet but researchers have built their own neural nets and we are looking into ways to incorporate some of the nn algorithms into Foldit
[2:56 PM] milkshakeiii: GPU utilization also has some things stewing, but more on the scientist side, it's a bit of a long path before it makes its way to Foldit, but there is a path
[2:57 PM] milkshakeiii: Rosetta update is still happening, indeed
[2:58 PM] Formula350: I like the science side, too! I figured offloading metrics might be easier in the short term heh
[3:00 PM] ZeroLeak7: thats sounds good
[3:00 PM] Formula350: Do you happen to know what the benefits are to us end-users, with the Rosetta update, in terms of what would be readily obvious to us?
[3:01 PM] milkshakeiii: more possible objectives would be the main thing
[3:01 PM] milkshakeiii: all of the fundamental stuff should stay the same
[3:02 PM] milkshakeiii: I believe it's the exact same score function
[3:02 PM] Formula350: Basically it'll be adding more "science" that will be untapped.
[3:06 PM] pc: hi
[3:00 PM] Formula350: Do you happen to know what the benefits are to us end-users, with the Rosetta update, in terms of what would be readily obvious to us?
[3:01 PM] milkshakeiii: more possible objectives would be the main thing
[3:01 PM] milkshakeiii: all of the fundamental stuff should stay the same
[3:02 PM] milkshakeiii: I believe it's the exact same score function
[3:02 PM] Formula350: Basically it'll be adding more "science" that will be untapped.
[3:06 PM] pc: hi
[3:06 PM] pc: oh I forget the office hour ^^
[3:06 PM] ZeroLeak7: it is now
[3:06 PM] Formula350: Hallo. It's still going, if you have any non-science Q's for him
[3:10 PM] Formula350: My other "compute pitch' is wayyy out there in terms of actual usefulness… Maybe seeing if recipes could be ported to soimething like Shaders, to process. (I'm not even sure if it's possible though)
[3:12 PM] pc: when we want to move a protein by keeping left mouse button pressed, there is something like a velocity. I makes time before starting move, and often it is moved too far. Is there a way to move protein slowly with precision ? I know we can cut, and move with the arrows and join again. but the residue is not ideal after that.
[3:12 PM] milkshakeiii: you mean the pull tool, right pc?
[3:13 PM] milkshakeiii: yeah i don't love the way it works either
[3:13 PM] pc: yes pull tool
[3:13 PM] milkshakeiii: the reason is because it's doing optimization as it runs
[3:13 PM] pc: sometimes we need to move with more precisions. In linker puzzles for exemple
[3:14 PM] Formula350: It's not horrible in small puzzles, but once you add in the Objectives, it does become clunky. Even pulling on Sidechains to change their rotamer has the same issue :\
[3:14 PM] milkshakeiii: it's definitely part of what i'm looking at in terms of performance
[3:14 PM] Formula350: Does the Wiggle Power impact the Pull tool?
[3:14 PM] Formula350: I know CI does…
[3:14 PM] milkshakeiii: i'm not sure
[3:15 PM] Formula350: If so, PC you could try setitng it to Low and see if that helps…
[3:15 PM] pc: of course "enable cut band" off is better is some cases too
[3:15 PM] pc: when on, it can help, but the force is maybe too strong sometimes.
[3:16 PM] Formula350: It won't impact a pull if ther aren't any cuts.
[3:16 PM] ZeroLeak7: yes you should do it off
[3:16 PM] Formula350: If there IS a cut, then yes, turning it off will free it to move easily, as it won't be "banded" to the rest of the protein.
[3:18 PM] robgee: milkshakeii could you mention some of the optimisations you have in mind, would that be stuff like reducing array writes, flattening nested loops , memoization, etc ?
[3:18 PM] Huubr: Speaking about moving with more precision than with a mouse: I would love to be able to select a residue in the Rama Map, and then move it using the cursor keys, pixel by pixel.
[3:18 PM] milkshakeiii: eliminating redundancy is the first thing
[3:19 PM] pc: In linker puzzles, when I want to close the last cut point, I never know if it is in a good position. It is hard to see.
[3:19 PM] milkshakeiii: as in eliminating unnecessary iterations, then eliminating unnecessary operations within a single iteration
[3:20 PM] milkshakeiii: the stuff you mentioned might come into play, just basic "fix the time complexity" techniques
[3:20 PM] milkshakeiii: and after all that making things more asynchronous in a way that doesn't mess with game flow
[3:21 PM] milkshakeiii: or hopefully improves game flow
[3:21 PM] milkshakeiii: it's usually stupid stuff, as it so often is in software engineering
[3:22 PM] milkshakeiii: for example last patch I stopped a bunch of things like deleting rubber bands from causing the whole score function to be rerun (deleting a rubber band can't affect the score)
[3:22 PM] ZeroLeak7: too many cooks make the soup too salty
[3:23 PM] Formula350: Too many chefs cook way too much spaghetti :P
[3:24 PM] robgee: sweet, love that kinda thing. optimising away a chunk of code feels so good :)
[3:28 PM] ZeroLeak7: we have a great dev team now thank you for your time @milkshakeiii and have a good start in the New Year! and stay well!
[3:29 PM] milkshakeiii: thank you for your positivity ZeroLeak!
[3:29 PM] milkshakeiii: happy new year to you!
[3:29 PM] ZeroLeak7: yes always positive thank you
[3:29 PM] pc: Foldit is a good project. I hope we will find medications
[3:29 PM] milkshaeiii: and happy new year to all folders 8)
[3:29 PM] Formula350: Indeed. It's nice to see Foldit getting some much needed under-the-hood love finally!
[3:30 PM] Dudit: @milkshakeiii Thank you very much!
[3:30 PM] robgee: Thanks milkshakeii, Happy New Year to you and the foldit team.
[3:30 PM] Formula350: Happy near years to you as well. :)
[3:31 PM] pc: there where some good improvements in Foldit this year ^^

[agcohn821]

[10:00 AM] horowsah:hi all- I’m here for an office hour! I’m usually involved in educational things and electron density puzzles, and on the rare occasion we try an RNA or DNA puzzle. I’m a biochemist, so I can also answer general biochemistry questions, but I’m not a true expert in protein design specifically, and definitely not the diseases that we’ve been having a lot of binder design puzzles for.
[10:00 AM] pc: hi horowsah
[10:02 AM] horowsah: i really should change my picture, come to think of it
[10:03 AM] pc: do you have some scientific feedbacks for linker puzzles ? We dont know what to try for now. We moslty make 5 helix and have good objectives score.
[10:03 AM] horowsah: so if i recall, we've mostly been asking for inflexible linkers, right?
[10:04 AM] pc: yes
[10:04 AM] horowsah: it's a tough challenge, because to be maximally effective, the linkers need to have good packing interactions
[10:04 AM] horowsah: on both sides, and ideally internally within the linker itself
[10:05 AM] spvincent: perhaps in that case we need more residues to work with
[10:05 AM] horowsah: how long have the linkers been so far?
[10:06 AM] spvincent: up to 250 but that includes the fixed bits.
[10:06 AM] horowsah: yeah, as you probably know, we try to keep puzzles under 250 because for a lot of people it slows down too much
[10:06 AM] horowsah: even with some parts fixed
[10:07 AM] horowsah: but there might be some things we could try to increase the size more
[10:07 AM] spvincent: theres something of an incompatibility between core filter values and interactions with the fixed elements
[10:08 AM] pc: In this 2 loops zone, maybe there is a problem to make the protein folding correctly. So sometimes I try to make bouds between the 2 loops. Is it correct, or we should avoid 2 loops beeing too close ?(edited)
[10:09 AM] SuperNova185:Different tech issue: I get site security alert on the portal url w EDGE browser (although the cert is valid) but not TOR or CHROME. can you explain why this is so?
[10:09 AM] horowsah: @spvincent i think you're right about a bit of an issue with the filters not being perfect for this sort of puzzle, it's something to improve on
[10:09 AM] horowsah: @pc: you see this sort of thing in nature, but usually it's very flexible and not all that stable. So it might not be the best strategy here.
[10:10 AM] horowsah: @SuperNova185 I really don't know, let me pass that one onto someone who might know that
[10:10 AM] pc: thanks
[10:10 AM] SuperNova185:Thanx
[10:11 AM] spvincent: when will we get some feedback as to which style of linker design looks most promising?
[10:12 AM] horowsah: i was just hearing that some are going to be experimentally tested in the near future, although don't quote me on that as i think i'm two steps removed from that actually happening.
[10:12 AM] spvincent: ok thats good.
[10:14 AM] pc: Here I made more bouds. Maybe it is better. If we resolve all buns in this loop zone, is it ok ?
[10:14 AM] susume: are there any plans for any ED puzzles soon?
[10:15 AM] SuperNova185:@susume: and DeNovo. I remember that you like those.
[10:15 AM] horowsah: @pc: getting rid of buns will help, but it still is usually better to try and get some regular secondary structure in there, like maybe to widen it out make make some sheets?
[10:15 AM] SuperNova185:@susume: DeNovo puzzles, that is.
[10:16 AM] horowsah: @Susume: sadly, no ED puzzles in the very near future, but I think it would be good to do that soon
[10:17 AM] susume: I had this idea foldit players were going to have a chance to re-interpret some problematic structures from the PDB using our ED skills, but I wonder now if alphafold might do that better (though I also sense there is not much appetite for doing it at all)
[10:17 AM] horowsah: good question- there's a few parts to that
[10:18 AM] horowsah: it seems like alphafold is really good at getting the overall fold right, but the details of sidechains and such it doesn't always get right
[10:18 AM] horowsah: so that's a spot where humans are likely still going to be needed
[10:18 AM] pc: The same zone but in different view. One of the other solution to avoid this can be let a little hole. But maybe it is worse. So maybe I should try other configuration to avoid this zone.
[10:18 AM] horowsah: and i doubt they have any immediate plans to get it working well with ED
[10:19 AM] SuperNova185:What skillset is required to be a DevPrev reviewer?
[10:19 AM] horowsah:@pc: yeah i see the issue there, it would be good to close that somehow, but that one seems tough
[10:19 AM] horowsah: @SuperNova185 mostly the tolerance to handle a lot of crashes and not get frustrated! There's probably more to it than that, but I'm not actually sure
[10:21 AM] spvincent: Slightly off topic, but I was wondering what is the future of Rosetta in this time of AlphaFold?
[10:22 AM] horowsah: @spvincent i consider it on topic, since it's pretty related to the science of Foldit. Again, it's a complicated answer. First, it's likely Rosetta will incorporate the advances of AlphaFold2 into itself within a couple of years
[10:22 AM] horowsah: and that will make certain tasks in Rosetta considerably more powerful; it will probably improve protein design both within Rosetta and for Foldit, for example
[10:24 AM] horowsah: @Supernova: in answer to your devprev question, anyone can play it who would like
[10:25 AM] susume: I believe a hybrid of Rosetta with trRosetta, plus a new quality assessor called DAN, was Baker lab's best prediction model in the last CASP - Dr Baker has recommended to his lab that they all use this model for structure prediction of designed proteins now, rather than Rosetta forward folding (which was the gold standard until now)
[10:25 AM] spvincent: Tx. I'm wondering how AlphaFold2 would manage if you gave it a protein with novel amino acids: it has nothing to train on in such a case
[10:25 AM] horowsah: @Susume: that's what I've heard as well
[10:25 AM] horowsah: @spvincent unnatural amino acids are something it would probably fail at
[10:27 AM] spvincent: So you would still have to use Molelular Mechanics-based prediction in such a case.
[10:27 AM] horowsah: yep, good old fashioned physics
[10:28 AM] spvincent: Well good old fashioned physics should have cracked the problem. Yet progress seemed to stall a decade or so ago looking at the Casp results
[10:28 AM] spvincent: Whats the missing ingredient?
[10:28 AM] pc: Rosetta have values built from natural proteins. So, is there a difference in precision with the proteins we create in foldit?
[10:29 AM] horowsah: quantum mechanics, for one. Also, imperfect modelling of water is a big issue
[10:30 AM] horowsah: @pc: novel folds, such as those built by foldit players that don't exist in nature, presumably could cross out of areas in which Rosetta has it's best coverage from a fragment sampling perspective, but I think it's probably a pretty small inaccuracy in the grand scheme of things
[10:30 AM] spvincent: Any plans to add QM to Rosetta? Perhaps limited to interesting parts of the protein, such as the active site of an enzyme?
[10:31 AM] horowsah: that's a good question, and i wouldn't be surprised if someone is working on it, but i haven't heard about it
[10:31 AM] spvincent: tx
[10:31 AM] horowsah: in general, hybrid qm/ molecular mechanics force fields are still a work in process
[10:32 AM] spvincent: A trio of scientists won the Nobel prize for work in this area some years ago IIRC
[10:32 AM] susume: trRosetta doesn't know anything about physics, but it makes Rosetta much more effective by narrowing the space Rosetta has to search for (physics-based) energy minima - I expect alphafold 2 will have the same effect only much more so
[10:33 AM] horowsah: presumably it will, and the two will likely work together well
[10:34 AM] susume: there are people thinking about how to add trRosetta to foldit, and there are people at Baker lab planning to build the next generation of trRosetta based on alphfold2, so eventually that will hopefully also come to foldit
[10:34 AM] horowsah: …but it might take awhile
[10:35 AM] spvincent: Have DeepMind actually published their results yet?
[10:36 AM] horowsah:
they published on alphafold1, but i would guess alphafold2 will probably not be published for a year yet
[10:36 AM] spvincent: I wonder if there'll be an AlphaFold3
[10:36 AM] horowsah: i hope so!
[10:36 AM] pc: When Mers puzzles lab tests will be done (and when we will have results ^^)
[10:37 AM] horowsah: i've heard some chatter about it, but they are apparently not easy, so i don't know the eta on those
[10:39 AM] susume: unrelated to foldit - is it true that imaging proteins with structural metal ions is difficult with x-ray crystallogrpahy, and do you know if cryo EM is making that easier?
[10:39 AM] horowsah: actually, i published a paper on that recently
[10:39 AM] pc: if mers puzzle worked, that means maybe the current metrics are goods ^^(edited)
[10:39 AM] susume: oo, link pls?
[10:40 AM] horowsah: https://journals.iucr.org/d/issues/2019/12/00/wa5123/
Acta Crystallographica Section D
Identifying dynamic, partially occupied residues using anomalous sc…
Structural studies of partially occupied, heterogeneous protein systems using crystallography are difficult. Here, methods are presented for detecting these states in crystals.
[10:40 AM] pc: nice
[10:40 AM] horowsah: crystallography is actually really good at identifying metals… if you have the right beamline
[10:40 AM] horowsah: cryo-em probably won't be as good at it for a long time
[10:41 AM] horowsah: @pc: we can hope, but the metrics are likely going to be improving still in the future
[10:42 AM] susume: ty, will definitely read! I'm taking a class on metals in biomolecules
[10:42 AM] horowsah: back to the question of imperfections in our ability to computationally model proteins: metals are near the top of the list
[10:43 AM] horowsah: there's a reason you don't see too many of them in foldit puzzles
[10:45 AM] spvincent: What are the issues?
[10:46 AM] horowsah: ok, take this with a grain of salt since i'm not an inorganic chemist, but metals have a way of binding things that are somewhat in between what we think of as covalent bonds and weak bonds that you manipulate in foldit, for example
[10:46 AM] horowsah: and classical physics has a tough time handling it for some reason that i'm not entirely sure of
[10:47 AM] horowsah: so usually metals are handled in a way that is pretty ishy
[10:47 AM] horowsah: and even our quantum mechanics models don't do a great job with a lot of metals
[10:48 AM] horowsah: @Susume: that paper only makes brief mention of metals, but the concept is the same as to what we were looking at; metals have distinct edges for anomalous scattering that can be picked up by x-ray crystallography and not that many other techniques
[10:51 AM] spvincent: It used to be the case that there was a bit of question mark hanging over structures obtained with X-ray crystallography: was the crystal structure the same as the one in solution.
[10:51 AM] spvincent: Does this still persist?
[10:51 AM] horowsah: people like to dredge up that argument from time to time but it really only applies in rare cases
[10:52 AM] horowsah: but if you want to see a fun one:
[10:52 AM] horowsah: https://www.pnas.org/content/99/5/2806
Crystal structure of an antiparallel DNA fragment with Hoogsteen ba…
We report here an alternative double-helical structure of the DNA molecule. It has been found in the d(ATABrUAT) and d(ATATAT) sequences by single-crystal x-ray crystallography. This sequence is found not only in TATA boxes, but also in other regulatory regions of DNA. Bases of the two antiparallel strands form Hoogsteen pairs, with adenines in …
[10:52 AM] horowsah: for the most part, crystallography really does show what happens in solution
[10:53 AM] horowsah: that paper is an entire dna helix in which hoogsteen base pairing happens instead of watson crick base pairing
[10:53 AM] horowsah: and while hoogsteen pairing is thought to occur rarely (like 1% of the time or less), an entire helix doing it simultaneously is probably an artifact of crystallography
[10:54 AM] spvincent: tx I'll look at it later. gtg: thanks for answering questions horowsah.
[10:54 AM] horowsah: np
[10:57 AM] susume: since @Formula350 is not here I will ask his signature question - is there anything you as a scientist wish that foldit players would do more of, or anything you wish we would do less of?
[10:58 AM] pc: In Foldit, is it better to play for points, and try to find the metrics limits, or to think about science, and make more viable solution but with less points ?
[10:59 AM] horowsah: so my answer to both of those questions is: it's best to try for novel solutions that are outside the box. Sometimes this makes it tougher to get the high score, but it's more scientifically valuable to be unique in your approach and solutions
[10:59 AM] pc: thanks
[11:00 AM] horowsah: thanks all, i have to run!

[agcohn821]

[3:07 PM] scriren: How is everyone doing today?
[3:09 PM] HuubR:I'm fine, thanks. A bit surprised about the office hour; I did not see the announcement, but that is actually a pleasant surprise.
[3:09 PM] HuubR:What will be the general subject?
[3:13 PM] scriren:
Hi HuubR so a little bit about myself. I am a software developer working on Foldit with the Meiler Lab in Nashville TN. I primarily work on Small Molecule design. I know there haven't been any small molecule puzzles lately, but we have a few things in the works and are hoping to get them out sooner rather than later
[3:14 PM] HuubR:Small molecule, as in Ligand Design? The only one I remember was 1955 (July last year), but then again I have only been folding since April :-)
[3:15 PM] Susume: cool! I can't remember, are the small molecules all organic, or are there other groups/ions available?
[3:15 PM] scriren: Yes indeed Ligand Design
[3:15 PM] spvincent: What kind of things Sciren? There have been some occasional logand design puzzles in the past but they've always been a bit buggy.
[3:16 PM] Susume: you are working on the pick-from-column-A, pick-from-column-B interface, right?
[3:17 PM] scriren: To your first question Susume we are focusing mostly on organic ligands.
[3:18 PM] scriren: Yes so I have been working on the Reaction Design Tool. This is the tool where you have the choice of partial structures to choose from to create your ligand.
[3:18 PM] HuubR:I think it would be great when we can work on parts of a ligand just as we do on individual residues (Selection Interface). What I remember from Puzzle 1955, the ligand was one big segment, although we composed it of three reagents.
[3:19 PM] scriren: The purpose of this is to help players create synthetically viable compounds that can be created in the lab.
[3:20 PM] scriren: How so HuubR? I'm not sure I followed.(edited)
[3:22 PM] scriren: Oh and to your question @spvincent I'm working to update Reaction Design Tool to help fix some of the known bugs as well as updates to the tutorial.
[3:23 PM] spvincent: tx
[3:24 PM] HuubR: Well, forgive me for being direct, but I think a lot of players did not like the three column user interface. My idea would be that a ligand is composed of three segments, and you can replace one (or more) of these segments by other reagents, just as you can mutate single residues. Q1: Does that make sense? Q2: is it possible to do that in the program?
[3:24 PM] scriren: I'm also working on an updated version of the Ligand Design Tool that takes a more atomistic approach. Where you can create your ligand (or modify it) by changing the atoms and bonds.
[3:25 PM] HuubR:That sounds even better than my suggestion :-)
[3:26 PM] Josh: you should see the mock-ups HuubR. I've seen them and Sciren is doing some amazing UI work
[3:26 PM] spvincent: Hasn't there been a lot of computational chemistry done over the years on this? Can Foldit players really improve on massive computational efforts such as OpenPandemic at the WCG?
[3:26 PM] Josh: he's taking Foldit in a beautiful direction
[3:27 PM] scriren: That's great to hear! No worries about the directness. I want to know how players use the tools provided and how to make them better!
[3:28 PM] scriren: Awe thank you very much @josh Its also great that you mention this because I want to share with the everyone the design for the new builder.
[3:28 PM] Skippysk8sIRC: I felt like a kid with a chem set in grade school I had no idea what order the selected pieces would assemble, so it felt random to me. My lack of organic chem probably doesn't help
[3:30 PM] scriren: I wanted to share this with you all, but I need to make clarify that this tool is under much revision and this will almost certainly not be the final iteration.
[3:30 PM] scriren: but this is the direction that I am trying to take the Ligand Design Builder
[3:32 PM] HuubR:Cool! It looks like a remote control. Where is the "detonate" button? :P
[3:32 PM] Skippysk8s: I can detonate anything .o.
[3:33 PM] HuubR:lol
[3:33 PM] scriren: @spvincent I think that Foldit players do have a lot to contribute especially when you take the spacial recognition abilities players excel at and couple them with the ligand-protein interface.
[3:33 PM] scriren: Haha! I haven't added that one yet, but like I said this is subject to change so there is always time to add it in
[3:35 PM] spvincent: It will be nice to have some variety in the puzzle types. Look forward to trying it: any idea when we'll get the first one?
[3:35 PM] scriren: My hope is, once the design is in a more solidified state, to open a sandbox style ligand puzzle as well
[3:37 PM] HuubR:Can you give any indication about the time frame for that?
[3:38 PM] scriren: What I can say is that I am working on getting this into your hands as soon as possible. I know that that isn't always the most satisfying of answers, but I want to make certain that you all have the best experience possible.
[3:40 PM] HuubR:It's probably not the answer we hoped for, but it's an honest answer, and therefore a good one.
[3:42 PM] scriren: Thank you. I very much want to see all the awesome science that can be done with these tools in your hands, so as a developer I am very excited to get this to you when it is ready.
[3:42 PM] Susume: I really like the interface - sleek, easy to understand (much better than the actual chemistry!)
[3:43 PM] scriren: I'm a gamer as well so I also understand what the anticipation is like.
[3:44 PM] HuubR:Just a first remark after admiring these beautiful pictures: on SideOne, I would expect N, O, and F in that order (7, 8, 9). Is there a specific reason to put O first?
[3:44 PM] scriren: Thank you @Susume the hope with this type of interface to get players up and running as intuitively as possible.
[3:45 PM] Skippysk8s: I think some of my problem with the last puzzle was an inability to control the orientation of the ligand . I'd end up with the wrong end trying to connect
[3:46 PM] Susume: F, Cl, I and Br (the halides) all act similarly in a molecule, so they can be swapped out, makes sense to me they are in a box together
[3:46 PM] HuubR: Yep, agree
[3:46 PM] pc: oh there is a office hour
[3:46 PM] pc: hi
[3:47 PM] scriren: HuubR I put O on the left hand side because I found myself using it more often and it was an ease of use decision.
[3:47 PM] scriren: You are correct @Susume !
[3:47 PM] scriren: Yes indeed! Hello @pc !
[3:47 PM] Susume: oh I see what you mean, yeah maybe N should go above P (they both like 3 bonds) and O above S (they like 2 bonds)
[3:48 PM] HuubR:Like in the periodic table.
[3:48 PM] scriren: That's actually a really good point and one that I can look into!
[3:49 PM] spvincent: There was a previous incarnation of the small ligand design tool that looked something like this. I had problems using it: I was never sure what it was that I was creating.
[3:49 PM] scriren: One of the original concepts that I worked on was arranged more like the periodic table, but to do the groups as @Susume mentioned, it was a little troubling to look at for anyone when a chemistry background.
[3:50 PM] Susume: when we get the heavy atoms wher we want them, can we fill all empty spots with H with one click?
[3:50 PM] pc: I imagine new molecule puzzle, with contextual menu to gradualy add atoms, keeping bouds in current molecule in game.(edited)
[3:50 PM] scriren: That is a use case that I haven't checkout out, but I can look into it.
[3:51 PM] Dudit: @Sciren I think the Objectives (Filters) UI/UX in Foldit needs to be improved (like the Formula350 suggested)
[3:53 PM] scriren: I think I see what you are saying pc. You are talking about just creating a molecule by growing it piece by piece?
[3:53 PM] pc: yes, like in design puzzles when we add residues
[3:54 PM] scriren: @Dudit I agree and I can assure you that it is being worked on as well.
[3:54 PM] spvincent: It might be too easy to create implausible molecules by just successively adding atoms.
[3:55 PM] scriren: Indeed, this is one of the issues with this approach.
[3:56 PM] HuubR:Would it be feasible to come up with a scoring system that tells you how plausible a certain molecule is?
[3:56 PM] pc: I imagine that there is constrains, so a contextual menu can show what we can add each time or not
[3:56 PM] scriren: Work is being done as well to help guide players to creating synthetically viable molecules. It isn't easily done, but we are working on it.
[3:57 PM] scriren: It's the entire premise behind the Reaction Design Tool, and one of the reasons I am working to improve it.
[3:58 PM] HuubR:Sounds like quite a challenge. The possibilities are, well, almost endless.
[3:59 PM] pc: In last puzzle I tested, each time I wanted to modify my molecule, I losed the position, rotation, and all bonds I have succesdeed to do.
[3:59 PM] scriren: Is this the Reaction Design Puzzle?
[3:59 PM] pc: yes
[4:00 PM] HuubR:@pc, I agree. It was not possible to leave the ligand in place and modify just one end of it. At least, I did not manage to do that.
[4:00 PM] jeff101: it would help to have LUA tools to list atom types for each atom in the ligand (C, N, O, Cl, etc) and perhaps even different types of C, N, and O atoms based on how many other atoms they bind to
[4:00 PM] scriren: Yes positional alignment is one of the issues being worked on.
[4:01 PM] scriren: That is something I can look into jeff101
[4:01 PM] scriren: Sort of a labeling system outside of the color scheme?
[4:02 PM] jeff101: like carbons can be part of CH3 or C=O
[4:02 PM] jeff101: O can be part of C=O or -OH
[4:03 PM] jeff101: carbons can be part of rings
[4:03 PM] pc: There were this menu. And the current molecules used in game where not selected when we open this menu. So we couldnt know what other molecule to choose, because we didnt know what where the current.
[4:04 PM] scriren: Oh I see, you mean something like listing functional groups. I can look into see if there is a good way to list common groups to help players build.
[4:05 PM] HuubR: @pc, scroll back to 23:31h (CET). The new user interface is awesome!
[4:05 PM] scriren: Ah yes, there should be an update soon @pc fixing the selection issue. A little "behind the scenes" info on that bug. When the tool is first loaded a selection was being made that was not being highlighted. I was able to track it down and correct it.
[4:05 PM] pc: cool
[4:06 PM] BootsMcGraw:It was very difficult for me to see which choices were selected when using the tool.
[4:07 PM] pc: the idea is that we cant to test multiple configurations like when we change some amino acids one by one to make some bouds (but I know we can use "mutate" too ^^)
[4:08 PM] BootsMcGraw:IMAGE: http://fold.it/portal/files/chatimg/irc_82314_1612566502.png

[4:08 PM] BootsMcGraw:IMAGE: http://fold.it/portal/files/chatimg/irc_82314_1612566519.png

[4:08 PM] scriren: I see what you are saying BootsMcGraw, the highlight isn't dynamic enough. I can place that within a future update. I'm working on getting a labeling system to you as well and I think that should shore up some confusion.
[4:08 PM] BootsMcGraw:IMAGE: http://fold.it/portal/files/chatimg/irc_82314_1612566527.png
[4:08 PM] pc: and maybe a "mutate" tool can by possible too to chose and add the best molecule extension at a current point
[4:08 PM] jeff101: I wrote at least one feedback about the last version of the REaction Design Tool. Some changes were less radical than the ones you are proposing in this office hour. One example is to keep the columns of reactacts but number them within the tool so it is easier to tell your teammates which reactant you chose for each column
[4:09 PM] jeff101: also to have LUA commands that would list which reactants were chosen from each column
[4:09 PM] BootsMcGraw:Thank you, Sciren, on behalf of the colorblind contigent.
[4:09 PM] scriren: Yes indeed jeff101 that is one that I am working on currently, and will have it out as soon as possible.
[4:10 PM] scriren: Happy to help BootsMcGraw!
[4:12 PM] scriren: With that I am going to have to step out folks. Thank you so so much for all of your time and wonderful feedback. You all make this community amazing! I very much look forward to speaking with you all in the future! Happy Folding Everyone!
[4:13 PM] jeff101: when you tab on a designed ligand, it could list the formula for it like C6H5Cl
[4:13 PM] Skippysk8s: thanks
[4:13 PM] HuubR:Thanks, @Sciren!
[4:13 PM] Dudit: Thank you @Sciren

[agcohn821]

[11:01 AM] bkoep: Hi @here! Welcome to office hours
[11:01 AM] bkoep: I'm bkoep, a scientist on the Foldit team
[11:01 AM] bletchleyPark:Hi bkoep
[11:01 AM] Dudit: Hello @bkoep
[11:01 AM] pc: hi
[11:01 AM] spvincent:hi
[11:01 AM] alcor29: Good morning.
[11:01 AM] bkoep: Here to talk about all things Foldit, and especially protein design
[11:02 AM] pc: @bkoep: When you study the solutions shared by the players, are repeated errors that you see a problem ? Or are you looking for a solution to give them feedback if needed (in a practical way of course, like a global feedback ? ).(edited)
[11:03 AM] Dudit: @bkoep Will there in the future a Foldit Designable Linker puzzle that is linking three or more proteins?
[11:03 AM] bletchleyPark:@bkoep: given the pandemic, I would welcome all focus on improving effectiveness of vaccines.
[11:04 AM] alcor29: How is the therapeutic LCB1 doing in trials?
@pc
@bkoep: When you study the solutions shared by the players, are repeated errors that you see a problem ? Or are you looking for a solution to give them feedback if needed (in a practical way of course, like a global feedback ? ).(edited)
[11:05 AM] bkoep: I do try to keep an eye out for common or repeated problems, because these seem to indicate problems with the way we are framing a challenge, or problems with the way we are scoring solutions
[11:06 AM] bkoep: One example of this is would be "triangular" H-bond networks in symmetric trimer designs, where three SER (or THR, TYR) form a tight equilateral triangle around the symmetric axis
[11:07 AM] pc: Thanks. the idea is therefore to make Foldit more intuitive.
[11:07 AM] bkoep: This was really popular among a lot of player designs, and we were able to adjust the H-bond scoring to discourage this
@pc Thanks. the idea is therefore to make Foldit more intuitive.
[11:08 AM] bkoep: Yes, exactly. One of the primary goals for the Foldit scientists is to make protein science as approachable as possible.
@Dudit
@bkoep: Will there in the future a Foldit Designable Linker puzzle that is linking three or more proteins?
[11:09 AM] bkoep: It's possible we could run a linker puzzle to connect 3+ domains, but unlikely.
[11:10 AM] bkoep: The main concern is puzzle size. The more residues that need to be modeled (and rendered) in a Foldit puzzle, the more demanding the puzzle will be for computational resources. A puzzle with three domains and two linkers might be pretty slow on a regular desktop.
[11:11 AM] bkoep: Also, that's a problem that I think we can easily split up into sub-problems
[11:12 AM] bkoep: If we get a bunch of solutions linking domain A -> B and in another puzzle we get solutions linking B -> C, then we can probably combine those solutions into A -> B -> C very straightforwardly. No need to model the final product in a Foldit puzzle
[11:13 AM] pc: @bkoep: Are current linker solutions good ? Some solutions had almost or all the objectives, what is missing to improve further?
bletchleyPark: @bkoep: given the pandemic, I would welcome all focus on improving effectiveness of vaccines.
[11:14 AM] bkoep: Yes! There is a lot going on in vaccine development research right now
[11:14 AM] bkoep: However, there aren't a lot of obvious ways that Foldit can be applied to the problem
[11:15 AM] bkoep: Vaccine effectiveness usually boils down to: finding out the best way to display an antigen to the immune system
[11:16 AM] bkoep: Most of that work involves working with known, well-characterized proteins (the antigens). There's not much of a role for protein structure prediction or protein design
[11:17 AM] bkoep: But there are still some opportunities to help indirectly
[11:17 AM] bletchleyPark: In what way do our binders fit in then ?
[11:18 AM] bkoep: For example, we might be able to use protein design to help modulate the immune response in the presence of a vaccine. A couple years ago we attempted to design binders for CR2; it's thought that controlling the activity of this cell surface receptor could improve immunological response to a vaccine
bletchleyPark: In what way do our binders fit in then ?
[11:19 AM] bkoep: Most of our binder design projects are attempts to develop a therapeutic
[11:20 AM] bkoep: This kind of therapeutic would be more like a drug that you could administer, to directly mediate the effects of a disease. In contrast, a vaccine is a strictly prophylactic formula that trains the immune system to handle future infections on its own
[11:21 AM] bletchleyPark:Sounds like a great way to contribute, are we on the right track with our therapeutic designs ?
bletchleyPark: Sounds like a great way to contribute, are we on the right track with our therapeutic designs ?
[11:23 AM] bkoep: Yes, I think so. Protein binder design is a very difficult problem, and we have only just started to tackle it in Foldit so we don't have much data yet
[11:23 AM] bkoep: But our computational analysis of Foldit player-designed binders suggests that the designs are up to snuff
[11:24 AM] pc: I suppose if Mers puzzle results are good, that would be decisive for this kind of puzzle ?
[11:24 AM] bkoep: What's left is to try and boost throughput (more designs) and to get experimental data from the lab
alcor29: How is the therapeutic LCB1 doing in trials?
[11:25 AM] bkoep: I'm sorry I don't have any inside info on the LCB in vivo testing. The last I heard, it was going into ferrets and probably soon into non-human primates
[11:26 AM] alcor29: Thanks bkoep:
@pc
@bkoep: Are current linker solutions good ? Some solutions had almost or all the objectives, what is missing to improve further?
[11:26 AM] bkoep: Yes! I think the latest linker design results have looked really good! I'm hopeful that we'll be able to get some experimental data soon.
[11:26 AM] jeff101:I'm doing trimer puzzle 1960 and wish I could add or remove segments. Is that possible in trimer puzzles?
[11:27 AM] bkoep: We are still figuring out the best way to set up puzzles for the linker design puzzles, so that the best scoring designs have a core that spans the entire protein, and good packing throughout
[11:27 AM] alcor29: Yes I would like to delete some segs also.
[11:28 AM] alcor29: On the trimer.
jeff101:I'm doing trimer puzzle 1960 and wish I could add or remove segments. Is that possible in trimer puzzles?
[11:28 AM] bkoep: Haha yes, this has been on our wishlist for a long time, and you might expect it would be a trivial feature to add
[11:28 AM] spvincent:Thats a long-standing issue that is apparently not easy to fix. Much as we would like it to be.
[11:29 AM] bkoep: Basically, the symmetry features in Foldit prohibit us from inserting/deleting residues as easily as we do in non-symmetric puzzles
[11:29 AM] HuubR: I really like to play Symmetry puzzles, but I have wondered what their scientific value is. Can you tell us something about that?
[11:30 AM] bkoep: There are other roundabout ways to implement the feature, but it will take some work and right now we are occupied with other development priorities
[11:30 AM] Dudit: @bkoep: Are there any plans to develop Windows Foldit 64bit ?
[11:30 AM] donuts554: hello
[11:30 AM] pc: @bkoep: Some linker solutions have 0 buns, all ideal loops, and a good core. What do you think we should focus on next ?
HuubR: I really like to play Symmetry puzzles, but I have wondered what their scientific value is. Can you tell us something about that?
[11:31 AM] bkoep: Of course! For me there are two primary motivations for symmetry puzzles
[11:31 AM] pc: I think straight helix is important too maybe
[11:32 AM] bkoep: The first motivation is that well-behaved symmetric proteins are extremely useful components for downstream applications
[11:32 AM] bkoep: Any kind of macro-scale protein materials (think synthetic silk, or membranes, or MOF-like bulk materials) will probably be built from symmetric components
[11:33 AM] HuubR: Can you define "well behaved"?
[11:33 AM] bkoep: So, the more well-behaved symmetric assemblies we have, the more easily it is for scientists to tackle those later projects
[11:33 AM] pc: interesting ^^
HuubR: Can you define "well behaved"?
[11:34 AM] bkoep: Yes, by "well-behaved" I mean things like: the protein is soluble at high concentrations; it self-assembles into the intended symmetry but not into off-target assemblies; it has a high melting temperature
[11:34 AM] bletchleyPark: @bkoep: I occasionally will run an optimization on an expired puzzle and post to scientists. Are you checking these 'late ' results ?
[11:35 AM] bkoep: The second motivation for symmetric design is that it can help us learn about designing protein-protein interfaces
[11:36 AM] HuubR: Maybe this will be another challenge for your programmers that is not a top priority, but would it be possible to make a Tetramer puzzle with two of the four subunits upside down? I see that in Wikipedia, in the 2nd picture for Tetrameric protein
[11:36 AM] HuubR: https://en.wikipedia.org/wiki/Tetrameric_protein
Tetrameric protein
A tetrameric protein is a protein with a quaternary structure of four subunits (tetrameric). Homotetramers have four identical subunits (such as glutathione S-transferase), and heterotetramers are complexes of different subunits. A tetramer can be assembled as dimer of dimers with two homodimer subunits (such as sorbitol dehydrogenase), or two h…
[11:36 AM] bkoep: The ultimate challenge is to be able to design a protein that can bind to an arbitrary target surface (this is "binder design"). In theory, it should be easier to design two proteins that bind to one another (or a protein that binds to itself), because we have control over both sides of the interface
[11:37 AM] bkoep: So we can focus on the low-hanging fruit (symmetric design) and still gain knowledge that will be useful for more difficult problems in the future (binder design)
[11:38 AM] HuubR: Thanks, that makes it a lot clearer!
@Dudit
@bkoep: Are there any plans to develop Windows Foldit 64bit ?
[11:38 AM] bkoep: I'm sorry, I'm not sure about plans to develop for 64-bit Windows. Are you unable to run the 32-bit app?
[11:39 AM] bletchleyPark:64-bit versions may handle the memory limitations better
[11:39 AM] alcor29: Is there ever a way for us to get to know something of what knowledge you are gaining when you look at our symmetric designs?
[11:39 AM] donuts554: Why is sulfur a nonpolar atom?
@pc
@bkoep: Some linker solutions have 0 buns, all ideal loops, and a good core. What do you think we should focus on next ?
[11:41 AM] bkoep: I think those are the biggest challenges for the linker solutions. We just need to make sure that a "good core" according to the Core Existence Objective is the same as a "good core" according to the protein engineers. These seem to be in agreement in the most recent puzzles, but it is still sometimes possible to satisfy the Objective and still have problematic "hinge" regions in the protein that allow movement of the linked domains
bletchleyPark: @bkoep: I occasionally will run an optimization on an expired puzzle and post to scientists. Are you checking these 'late ' results ?
[11:42 AM] bkoep: No, we do not look at puzzle solutions that come in after a puzzle expires
[11:43 AM] bletchleyPark:That is a pity
[11:43 AM] bletchleyPark:and good to know
[11:43 AM] jeff101:are there plans to change the ideal loops filter so a 3-residue non-ideal loop gives less penalty than a 12-residue one?
[11:43 AM] pc: thanks. And if we "share to scientist" some minutes after puzzle expire, it is not too late ?
HuubR: Maybe this will be another challenge for your programmers that is not a top priority, but would it be possible to make a Tetramer puzzle with two of the four subunits upside down? I see that in Wikipedia, in the 2nd picture for Tetrameric protein
[11:43 AM] bkoep: Yes, absolutely! In addition to cyclic tetramers (C4 symmetry), we can design dihedral tetramers (D2 symmetry) in Foldit
[11:44 AM] HuubR: :-)
[11:44 AM] jeff101:how about tetragonal tetramers (like 4-side dice in Dungeons & Dragons) ?
[11:44 AM] bkoep: I suspect D2 assemblies will be slightly more difficult to design, but it is certainly something we can try
[11:45 AM] pc: Solutions in design puzzles have more points for proteins with max residue count. Do you think the "residue count Objective" should be adapted ?
@alcor29
Is there ever a way for us to get to know something of what knowledge you are gaining when you look at our symmetric designs?
[11:47 AM] bkoep: Most of the "gaining knowledge" happens very late in a project, after we've met some success and we can step back and look at all of our experiments together
[11:48 AM] bkoep: (Sometimes we do not realize a conclusion until after we sit down and start to write a paper about our experiments)
[11:49 AM] alcor29: k
donuts554: Why is sulfur a nonpolar atom?
[11:51 AM] bkoep: Sulfur is on the edge of polar vs. non-polar (chemical polarity is more of a spectrum than a hard classification). Sulfur is slightly polar and can make weak hydrogen bonds, but it can also be buried without too much penalty
jeff101:are there plans to change the ideal loops filter so a 3-residue non-ideal loop gives less penalty than a 12-residue one?
[11:51 AM] bkoep: Probably not. In my eyes, a 3-residue non-ideal loop and a 12-residue non-ideal loop are equally problematic.
11:53 AM] jeff101:but for recipes, if the 3-residue ones have less penalty, it helps the recipe gradually reduce the # of residues involved in non-ideal loops
@pc thanks. And if we "share to scientist" some minutes after puzzle expire, it is not too late ?
[11:53 AM] bkoep: Please share your solutions before a puzzle ends. If you share a solution after the puzzle expires, we will probably not see it.
[11:53 AM] donuts554: How strong are the hydrogen bonds involving sulfur compared to pi stacking?
[11:54 AM] jeff101:now if a recipe fixes the middle of a 12-residue bad loop, it could end with two 3-residue bad loops, which gives a larger penalty
[11:53 AM] jeff101:but for recipes, if the 3-residue ones have less penalty, it helps the recipe gradually reduce the # of residues involved in non-ideal loops
@pc thanks. And if we "share to scientist" some minutes after puzzle expire, it is not too late ?
[11:53 AM] bkoep: Please share your solutions before a puzzle ends. If you share a solution after the puzzle expires, we will probably not see it.
[11:53 AM] donuts554: How strong are the hydrogen bonds involving sulfur compared to pi stacking?
[11:54 AM] jeff101:now if a recipe fixes the middle of a 12-residue bad loop, it could end with two 3-residue bad loops, which gives a larger penalty
[11:54 AM] Susume: jeff makes a good point, sometimes the importance of the score is that it provides something for recipes to optimize on, not just that it uniformly discourages undesired attributes
[11:55 AM] jeff101:if you have the penalty depend on the # of residues in non-ideal loops, recipes can gradually fix them
[11:55 AM] bletchleyPark:I'm off and will read up in the log, thank you for answering questions in this office hours bkoep:
[11:55 AM] bletchleyPark: afk
jeff101: how about tetragonal tetramers (like 4-side dice in Dungeons & Dragons) ?
[11:55 AM] bkoep: Yes, tetrahedral symmetry is also possible, but each "side" of the tetrahedron must have perfect C3 cyclic symmetry. So, in practice, you need at least 12 protein chains to make a tetrahedron. That's getting rather large for Foldit, so I would not expect a tetrahedral design puzzle anytime soon
@pc Solutions in design puzzles have more points for proteins with max residue count. Do you think the "residue count Objective" should be adapted ?
[11:56 AM] bkoep: Yes, we may need to recalibrate the score penalty for residue insertions
donuts554: How strong are the hydrogen bonds involving sulfur compared to pi stacking?
[11:58 AM] bkoep: I'm sorry I don't know this off the top of my head
[11:58 AM] jeff101:if we could read with LUA which residues are parts of non-ideal loops, our recipes might be able to fix such loops better
[11:59 AM] alcor29: Keeping inmind that fixing loops has, I think, become a real bugaboo for many folders.
[12:00 PM] jeff101:a recipe could count the # of continous residues in a non-ideal loop and then rebuild/remix the entire loop at once
[12:03 PM] pc: Rosetta score is good for the folding prediction, but in design puzzles, maybe it works less well for compare proteins which do not have the same number of residue. And in Design puzzles, virus and Binder are calculated by Rosetta score as if they were the same protein. Will the Objective always be enough to adjust score, or do you think you have an adapted new Rosetta score system ?
jeff101:if we could read with LUA which residues are parts of non-ideal loops, our recipes might be able to fix such loops better
[12:03 PM] bkoep: I think this is probably the ideal solution. We want the score function to reflect our scientific goals as accurately as possible. I don't like the idea of adjusting the score function to make a method more convenient; rather, I think the method should be adjusted in response to the score function
[12:05 PM] jeff101:I heard that some Foldit players are helping Josh code changes to Foldit (not just recipe writing, but writing new LUA functions, for example). Is this true?
@alcor29 Keeping inmind that fixing loops has, I think, become a real bugaboo for many folders.
[12:07 PM] bkoep: In a sense, this is a little reassuring. Loop closure is a huge problem everywhere in de novo protein design. It is a truly difficult challenge, so we should expect some difficulty in Foldit. Of course we would like to make it easier, but if Foldit players had no difficulty whatsoever with loop closure then I might become suspicious that we are modeling something incorrectly
[12:09 PM] jeff101: please make a LUA command that lists whether a segment is part of a non-ideal loop. then we can do trickier things in recipes to fix bad loops.
[12:09 PM] pc: it is very hard to join the last cutted point in linker puzzles yes
[12:10 PM] alcor29: Agreed if you are willing to give up quite a few designs. I try every possible approach to solving them and after all is done , I just accept the bad loops and give up. I gather though that quite a few others are having the same reactions.
[12:10 PM] alcor29: For my part I'm ok with it since I can only do what I can do.
@pc Rosetta score is good for the folding prediction, but in design puzzles, maybe it works less well for compare proteins which do not have the same number of residue. And in Design puzzles, virus and Binder are calculated by Rosetta score as if they were the same protein. Will the Objective always be enough to adjust score, or do you think you have an adapted new Rosetta score system ?
[12:11 PM] bkoep: The Rosetta score is supposed to reflect the energy of the system, so it is appropriate that two chains (e.g. binder and target) both contribute to the same score. But you are right that there are fundamental problems with the idea of optimizing score (energy) in a design puzzle
[12:11 PM] bkoep: (More about that in this blog post: https://fold.it/portal/node/2005623)
[12:11 PM] pc: thanks
[12:12 PM] bkoep: Actually, if we can manage it, we usually prefer to adjust the Rosetta score (instead of adding Objectives)
[12:12 PM] bkoep: We can model things more efficiently if everything is part of the Rosetta score function
[12:13 PM] alcor29: To rephrase I get that x number of designs will not be usefull. It just depends on how many casualties you are happy to accept.
[12:13 PM] bkoep: But we need for all terms of the Rosetta score function have to be differentiable and pairwise decomposable, and sometimes this is not possible for certain features
[12:14 PM] pc: the Rosetta score system in foldit, add sub-scores. But maybe there is synergy between some sub-scores, and it is better to multiply some and not just add ?
[12:15 PM] pc: and there is pairwise ok
[12:16 PM] HuubR: Thanks, bkoep:, for a very interesting Office Hour! Unfortunately, I have to go now.
[12:16 PM] HuubR: Looking forward to playing a D2 Tetramer puzzle :-)
jeff101:I heard that some Foldit players are helping Josh code changes to Foldit (not just recipe writing, but writing new LUA functions, for example). Is this true?
[12:17 PM] bkoep: Yes! We can share (most of) the Foldit source code to anyone who licenses Rosetta for non-commercial use (this is free to do). Contributing to the code is not easy, and since Foldit is not open-source we don't expect many people will want to contribute. But if you are interested, you should email mail.fold.it@gmail.com or send a PM to @josh
[12:20 PM] jeff101:thanks bkoep:
[12:20 PM] jeff101:soon will be the monthly video about Foldit
[12:20 PM] pc: is it finish ?
[12:20 PM] pc: so thanks ^^
[12:21 PM] bkoep: Okay everyone I have to split
[12:21 PM] bkoep: Thanks all for the excellent discussion!
[12:21 PM] jeff101:thanks for having this chat
[12:21 PM] alcor29: Tx bkoep:
[12:21 PM] Dudit: @bkoep: Thank you very much!
[12:21 PM] pc: Thanks for answers. always usefull for us.
[12:21 PM] pc: and for Foldit team I hope too.