Office Hours

[agcohn821]

[1:04 PM] jflat06: Alrighty, hello everyone, and welcome to office hours! I'm jflat06, a developer on Foldit. I work on the software side of things for both the client and server. I don't have anything specific to talk about today, so if you have any questions or topics, feel free to bring them up
[1:05 PM] spvincent: hi jflat. I'm wondering if you're making progress on fixing the problems with small molecule design
[1:09 PM] jflat06: While I don't work on that specifically, I know that those are actively being worked on. I think we have 1 fix waiting already, and more should be coming soon as well.
[1:09 PM] spvincent: ok, tx
[1:11 PM] jflat06: We have several developers who are dedicated specifically towards development on the small molecule design stuff
[1:11 PM] spvincent: i know from experience how hard it is to track down irreproducible bugs
[1:13 PM] jflat06: Yeah! It's definitely a nightmare sometimes
[1:14 PM] jflat06: We have been trying to add more debugging info over the years to help with these things, but it's always a tough problem
[1:15 PM] jflat06: One of the features for the new site that I'm really looking forward to is the ability to download and install any old version of the client, which will be amazing for enabling players to help track down bugs for us
[1:15 PM] pc: Hi. AlphaFold tool is very interesting and helpfull in puzzles. Will you add some additional informations in AlphaFold prediction, like confidence detail in each segments (with color for exemple) ?
[1:15 PM] jflat06: @pc definitely
[1:15 PM] jflat06: We have some things in mind, and we hope those features will be coming sooner than later
[1:16 PM] spvincent: Perhaps you could look into sending out versions which display more debug info: stack traces and the like.
[1:16 PM] pc: I saw this in a scientific software video (it is confidence color)
[1:16 PM] jflat06: There are some technical hurdles to overcome to communicate that info between the client/server
[1:17 PM] jflat06: Yup! The stuff we're looking at specifically is giving you confidence on a per-residue basis
[1:17 PM] LociOilingIRC: do we have any ideas on why sometimes the traceback info on windows is all zeroed out?
[1:17 PM] jflat06: You mean the crash stack traces?
[1:18 PM] jflat06: Unfortunately I'm not sure
[1:18 PM] LociOilingIRC: yes, seems like something must have zapped memory
[1:18 PM] LociOilingIRC: also sometimes we get a debug.txt after a crash, but often we don't
[1:19 PM] jflat06: Yeah, the crash reporter tries to get that stack trace, but some crashes are just 'segfaults', where the crash reporter never even gets a chance to grab that stack trace
[1:19 PM] jflat06: and unfortunately not much we can do in those cases
[1:20 PM] LociOilingIRC: ok, that makes sense
[1:20 PM] spvincent: Maybe there are better tools on other platforms
[1:21 PM] pc: Thanks. In foldit it is hard to visualize clashes and voids with details. to optimise packing we mostly use script and not enough hand folding. So maybe we need this kind of visualisation : (edited)
[1:23 PM] pc: with a volumetric cloud in a 3d texture
[1:26 PM] jflat06: I know we have the isosurface, but I realize that is a little… lacking
[1:29 PM] LociOilingIRC: just thought of this again, but would it be possible to have a protein design sandbox with alphafold?
[1:29 PM] spvincent: I was thinking the same thing!
[1:29 PM] LociOilingIRC: puzzle 2027 can be used up to a point, but it has a limit of 120 segments
[1:30 PM] pc: Do you know if players use AF in puzzles ? I saw some top score players forgot to use it. Maybe score system can take count about some protein analysis in shared solutions ?
[1:30 PM] LociOilingIRC: also, AF seems doubtful about some real proteins
[1:31 PM] pc: yes LociOilingIRC ^^. I experiment new proteins in a design puzzle round begin. But can be interesting in sandbox yes ^^.
[1:31 PM] LociOilingIRC: for example, I just submitted the zinc binding protein, and the prediction came back with only 60% confidence
[1:31 PM] spvincent: I don't think its a question of forgetting to use it.
[1:32 PM] LociOilingIRC: not forgetting, sometimes you just have to move on without AF
[1:33 PM] spvincent: I suppose at some point the AF score might get integrated into the Foldit score but i've no idea how that could reasonably be done.
[1:33 PM] jeff101:I've had solutions that start with good AF scores, but as they gain Foldit points, the AF confidence score goes down.
[1:33 PM] pc: sometimes we have to reduce points to get better confidence. Score system can be adapted maybe.
[1:33 PM] alcor29: Hi jflat06. would you know if foldit will work with Windows 11?
[1:33 PM] jflat06: I don't run the puzzles, but we might have some plans for that in the future with a setup like the competition we ran earlier this year where you try to get as many viable solutions as possible.
[1:33 PM] jflat06: @alcor29 I have no idea :x
[1:34 PM] jflat06: I don't think anyone on the team has upgraded yet
[1:34 PM] alcor29: k
[1:34 PM] alcor29: Tx.
[1:35 PM] pc: In binder puzzles when we adapt too mush our protein to target, the alone protein become less stable alone. it is normal. For the first time we can know when it happen with AF. But we can need an objective in game that tell us when out protein is to mush twisted. (edited)
[1:35 PM] jeff101:is AlphaFold based just on structures of monomeric proteins that contain no metals, ligands, or other cofactors and are not part of larger protein complexes?
[1:36 PM] jeff101:is AF based only on structures in the physiological pH range that most FOldit puzzles are done at?
[1:37 PM] pc: In last puzzle I maybe found a protein that keep his hight confidence during all round (91.5%). It is because I talked it to bko-ep this week ^^
[1:37 PM] jeff101:are any of the structures AF is based on membrane proteinns? are any in environments that favor disulfide bonds? are any in environments that disfavor disulfide bonds?
[1:38 PM] jflat06: I'm not sure about what training data was used to train alphafold
[1:38 PM] jflat06: I think the paper would be the best place to look up what training set they used
[1:40 PM] jeff101:it would be nice to be able to color residues by any subscore we choose. This would give a different perspective on things like packing, voids, and clashing.
[1:41 PM] jflat06: Yeah, that would be a cool feature. I know we have some plans for re-doing some of the visualizations, and that might be something we try and work in
[1:41 PM] jflat06: no timeline on that, though
[1:42 PM] jeff101:it would also be nice if we could access the # of neighbors Rosetta/Foldit uses to decide which residues are in the core
[1:42 PM] alcor29: Will any work on the new client be directed to improving the hypersensitivity of the move tool?
[1:44 PM] spvincent: this may be off topic a bit, but what will Rosetta be doing in this new era of AF and the Rosetta equivalent? looking at proteins with non-canonical amino acids: that kind of thing?
[1:45 PM] jeff101:can you cite a paper or web page that gives more details about how Rosetta/Foldit uses cones to find the # of nearest neighbors to a residue? I would guess that the cone angle and size depends on what amino acid a residue is. Things like glycine and alanine would have smaller cones than tryptophan for example. If we knew all the angles and distances involved, we might be able to code our own neare
[1:45 PM] jeff101:st neigbor calculator with LUA
[1:45 PM] jflat06: @alcor29 can you describe the problem more specifically? Is the rotation too sensitive?
[1:46 PM] jflat06: @spvincent I don't know all of what is happening on the Rosetta side of things, but I do know that there's a huge rush to figure out how to use and extend the machine learning algorithms, specifically with respect to protein design
[1:47 PM] jeff101:Rosetta@Home is still sending out jobs
[1:47 PM] jflat06: Even from a Foldit point of view, this is super exciting, because it gives us a chance to dramatically increase the odds that your designs will fold up properly by the time they make their way to us
[1:48 PM] pc: rotate around protein in symetric puzzles is not very easy. It is because it depends of protein position I suppose ? When protein is far it is less easy.
[1:48 PM] jflat06: @jeff101, I don't know much about the inner workings of the nearest neighbor calculations, unfortunately
[1:48 PM] jeff101:@pc do you use q to center the rotation on the protein. it helps
[1:48 PM] alcor29: Very hard to impossible to make a slight correction, especially with the gimble function. Can you do something akin to changing the gear ratio?
[1:48 PM] pc: thanks jeff ^^
[1:49 PM] jflat06: yeah I think I know what you're talking about. I think we could improve that, I always found it finicky, but I never actually had to do anything with it!
[1:49 PM] jflat06: I just had to test it… hah
[1:49 PM] spvincent: I can understand how the field has been turned upside down but isn't there a danger of relying too much on AF and abandoning Molecular Mechanics approaches? The latter should be much more general.
[1:50 PM] jeff101:I made a recipe called FindMonomers to help find monomers when they are far apart. It puts some bands between the monomers that you can follow from one monomer to another. It also adds labels to at least one monomer so you can find it using Notes Mode.
[1:50 PM] Vincera: Is there a way for bona fide, regular, independent FoldIt researchers without current instituitonal affiliation or gain access to full-text articles via UW? We often need to read beyond the abstracts and cannot afford 40$ a pop.
[1:51 PM] jflat06: I don't think anyone is abandoning it, but more thinking about when and where it can be integrated to improve existing techniques
[1:53 PM] spvincent: integrated into RosettaFold?
[1:53 PM] jflat06: @Vincera Generally, I'm not sure. I think any of the studies related to Foldit we pay to make open-access, though
[1:54 PM] jeff101:Vincera, if you know the title and authors, you can sometimes find pdf's of key articles doing a google search. Some authors post them on their own web pages.
[1:55 PM] Vincera: Scribd and Semantic Scholars offer pdfs but I don;t want someone's hijacked, stolen work.
[1:55 PM] jflat06: There are tons of different methodologies under constant research within the Rosetta community, and I think people are all asking themselves, "how does this fit into my research?"
[1:55 PM] spvincent: Also if you ask the authors for a copy saying who you are and why you want it they'll usually send you a copy.
[1:56 PM] jeff101:back in the day, authors would order hardcopy preprints or offprints that they could mail to colleagues
[1:56 PM] Vincera: Yes; but, as an American citizen, I cannot - for instance - directly request an Iranian article. This week I ahd to do a workaround and access via the Dutch researcher.
[1:57 PM] jflat06: Sometimes that might be, "can I just do this all with machine learning?". But more often I would expect people to ask, "What if this part of my research was a solved problem?"
[1:57 PM] Vincera: And some of the Caribbean Basin works are the most difficult.
[1:57 PM] Vincera: I read the French and Spanish research
[1:57 PM] Vincera: but cannot gain access to the researcehrs themselves often.
[1:58 PM] Vincera: Much of their work has not even made it into the databanks.
[1:58 PM] jflat06: For example, from a Foldit point of view, what if protein structure prediction was a solved problem? How do we redesign foldit with that in mind? We would want to have more of a focus on sequence manipulation and tools to match a sequence to a specific shape.
[1:59 PM] Vincera: I myself am published and had to pay for my own article.
[2:00 PM] spvincent: Perhaps but Foldit has been emphasizing protein design for years
[2:01 PM] alcor29: AF can predict monomers, but we have to change sequences to "bind" with other substances or symmetric copies. That's where I have a problem seeing where AF can help unless it is trained on millions of interfaces?\
[2:01 PM] pc: Foldit is more a "design software" than a "prediction software" now
[2:02 PM] pc: And I think in the future of Foldit we will have design puzzles with proteins that have "functions" (maybe)
[2:02 PM] pc: It will be very interesting ^^
[2:03 PM] jflat06: AF doesn't currently pick shapes that bind to other shapes. But having an oracle that tells you what shape a specific sequence folds up into is still enormously powerful for this task.
[2:03 PM] alcor29: Absolutely.
[2:04 PM] alcor29: Thanks jflat.
[2:05 PM] jflat06: That's never existed in the field before, and rather than viewing it as encroaching, I think most people in the field are beyond excited about how they can apply it to different problems.
[2:05 PM] spvincent: Maybe. I just have reservations about the way AF is sometimes portrayed in popular science websites as being the last word in protein structure prediction. It may be a major advance but ultimately it's pattern matching not science. Even if its the pinnacle of pattern matching
[2:06 PM] alcor29: No doubt. I have a feeling the interface problem will eventually be worked on.
[2:06 PM] jflat06: Sure - but a lot of science itself is pattern matching. Much of rosetta is based on statistical models and trying to emulate patterns.
[2:07 PM] jflat06: I certainly don't think that AF means the end of statistical of physical predictive modelling for protein folding.
[2:07 PM] spvincent: thats good
[2:07 PM] pc: AF give a very important feedback in design puzzles. But yes prediction is not resolved.
[2:07 PM] NinjaGreg: Seems to me what you'd want is a tool that allowed you to specify a fold, and have it tell you the sequence to get there.
[2:08 PM] jflat06: @NinjaGreg that's the holy grail
[2:08 PM] alcor29: That would be great!
[2:08 PM] NinjaGreg: Right now it's sort of "try this and see what happens".
[2:08 PM] pc: NinjaGreg : maybe it is possible yes, most of deep learning algo can be reversed (edited)
[2:09 PM] NinjaGreg: Think the AF people would look at that?
[2:10 PM] pc: https://fold.it/portal/node/2011976 : I made this topic for that @NinjaGreg ^^
[2:10 PM] jflat06: I'm sure they will at some point.
[2:10 PM] NinjaGreg: Hey, thanks pc!
[2:10 PM] jflat06: I'm pretty sure Rosetta people are
[2:10 PM] alcor29: You can have a library of sequences and corresponding folds, even the unconforming ones.
[2:11 PM] jflat06: Ok, I am going to go grab some lunch. Thank you everyone for coming!
[2:11 PM] alcor29: Thanks jflat
[2:11 PM] spvincent: tx jflat
[2:11 PM] NinjaGreg: Thanks for hosting, @jflat06!
[2:12 PM] pc: "bon appétit"
[2:12 PM] pc: thanks
[2:12 PM] Nicm25: thanks very much ^^.
[2:12 PM] jeff101:thanks jflat06

[agcohn821]

10:59 AM] rmoretti: Hello! rmoretti here for Office Hour. Let me know if you have any questions about Rosetta, Drug Design, the new UI (in devprev) or anything else.
[11:00 AM] cjddig: (he/him): Hello rmoretti!
[11:02 AM] HuubR: Hello @rmoretti :-)
[11:02 AM] alcor29: Good Morning moretti.
[11:03 AM] rmoretti: What do people think about the VHL drug design puzzles?
[11:03 AM] HuubR: Interesting type of puzzle!
[11:04 AM] HuubR: Totally different to what we are used to, though
[11:04 AM] cjddig: (he/him): It's interesting, and I'm enjoying it. But I want to change the direction of carbon chain of the ligand, but it is hard to do. Is there any effective way to do that?
[11:04 AM] pc: hi
[11:05 AM] Nicm25: nice new puzzle types, trying to find it unknown algorithm ;D
[11:06 AM] pc: New Ligand puzzle tools are a lot better than the previous. Off course there is some improvements possible ^^
[11:06 AM] alcor29: Same as cjddig:. We need more functionality. For example can we freeze just one carbon, etc.
[11:06 AM] rmoretti: To move the carbon chain, you can try pulling the ligand to try to get the end bits to move to a different location, but I realize that's not all that useful at the moment. We are thinking about adding a tool which will allow you to rotate bonds, but we don't have it fully implemented yet.
[11:06 AM] alcor29: Hard to manipulate the ligand.
[11:06 AM] HuubR: It feels like I am just trying random things (not having any background in chemistry, except in school, some 40odd years ago :-), to see what works, but I like it.
[11:07 AM] rmoretti: Finer atom-level control (e.g. freezing atoms) is something that would be nice, but it's a bit tough, as Foldit was set up to work mostly on a residue-level basis, not on an atom level.
[11:07 AM] pc: it is now easy to work on our ligand, keeping some bonds. But the may difficulty is to keep something easy to create in lab, and keep low clogp.
[11:09 AM] rmoretti: Striking that balance is what makes the small molecule design process tough (even for experts).
[11:10 AM] rmoretti: We're interested in giving you tools to make it easier, but we recognize we may still have a ways to go.
[11:10 AM] cjddig: (he/him): cLogP is one of the hardest part of playing the puzzle. If I put too much polar atoms, TPSA score goes down; if I remove them, cLogP and synthesizability score goes down.
[11:12 AM] cjddig: (he/him): By the way, how does foldit judge if the ligand is synthesizable? What makes ligand unable to be synthesized?
[11:14 AM] rmoretti: Synthesizablility is a little bit tricky, as often the only way to figure out if something is synthesizable is to show it to a medicinal chemist and ask them to come up with a synthesis scheme. For the synthesizability score, we use a published protocol which is based on looking at small segments of the ligand and ranking them based on how frequently they're seen in drug-like molecules. If a small section of the ligand isn't seen frequently, it means it likley isn't very synthesizable.
[11:16 AM] pc: @rmoretti A little improvement can be better tooltips. The real name of the atom, his main utility. I never understand F I Cl and Br utility for example.
[11:16 AM] rmoretti: That's what the "show" button of the Synthesizability objective is doing – it's showing those sub-sections of the ligand which aren't seen very often, and are thus contributing negatively.
[11:17 AM] rmoretti: @pc Good point – the tooltips could have the full element names. (That's a blind spot a chemist has - we know all the abbreviations.) – By the way, F is Fluorine, I is Iodine, Cl is Chlorine and Br is Bromine
[11:18 AM] pc: Thanks. and a little description on the utility like "interact with watter", or "usefull to create this.."
[11:19 AM] spvincent: I've yet to find a case adding/replacing with these atoms makes a positive difference.
[11:20 AM] HuubR: Has the synthesizability objective been changed after the first round? In p2059, my ligand was synthesizable, and I think I entered exactly the same formula in p2062, but then it was "very likely not synthesizable".
[11:20 AM] alcor29: Why do we begin with lots of Nitorgen, for example, since we will probably have to get rid of them to meet the amide etc. penalties/undesirable group requirements.
[11:20 AM] rmoretti: The "useful to create this" tooltip would be a bit harder, as often those sorts of things are highly contextual. – It's probably something that would need to be an external resource, sort of like how you learn when Lysine or Asparagine are useful in a protein.
[11:20 AM] Susume: (she/her): maybe a wiki page
[11:21 AM] rmoretti: @HuubR We did make the synthesizability objective a slight bit more stringent in the most recent round.
[11:21 AM] HuubR: :-)
[11:21 AM] cjddig: (he/him): By the way, when I try to put phosphorus, it always say "chemically not feasible." Also, I learned that nitrogen can make bonds up to 5, but it seems to be impossible. Why is that?
[11:22 AM] rmoretti: @alcor29 The molecule we're starting with in the VHL puzzle is a compound from the literature which already has a decent binding. Changing up the compound to meet the various objectives is sort of the goal of the puzzles.
[11:23 AM] spvincent: Broadly speaking, are you looking for a few high-scoring solutions or many solutions with more variety but which perhaps score less?
[11:23 AM] alcor29: In general we add and delete things kind of randomly and see if the score goes up. Maybe that's a good thing for you. Introducing the ignorant human element. Or, maybe we need more chemistry knowledge to do a better job?
[11:24 AM] HuubR: I'm with you, alcor29 :-)
[11:24 AM] rmoretti: Phosphorus has rather different requirements for the number of bonds, so typically if you just want to replace an existing atom with phosphorus it's not likely to work. Adding a phosphate group from the fragments panel is likely the best approach.
You don't typically see nitrogens with five bonds in drug-like molecules, so we've limited it to only having 4 bonds.
[11:26 AM] rmoretti: @spvincent We haven't yet dialed in the scoring function to get compounds which are interesting to show up as high scoring. That's somewhat anticipated in these early stages, but if you're looking for scientific impact, rather than score, trying a variety of different approaches to reach the stated goals of the puzzle (from the puzzle description) is probably the way to go.
[11:26 AM] pc: @rmoretti A little other improvement can be : when we select a new fragment, an UI can show us what fragements are more "likely synthesizable" for exemple.
[11:29 AM] rmoretti: @alcor29 More chemistry knowledge would probably be helpful, just like more biochemistry knowledge is helpful in the protein folding and design puzzles. Citizen science drug design is still in its very early stages, so we're still trying to figure out what's the best level of approach, and how much "expert" knowledge is needed.
We sort-of have the hope that the community can come together to discover what's useful and what's not in their considerations. But if there's something in particular we can do to help out, please let us know.
[11:30 AM] spvincent: You could fix the crashing :)
[11:31 AM] HuubR: I would guess that you adapted the synthesizability objective based on us non-chemists coming up with compounds that a chemist would never think of?
[11:32 AM] alcor29: Yes, it's crashy again.
[11:32 AM] cjddig: (he/him): Is sulfur counted as polar atom or non-polar atom? h TPSA and cLogP highlights sulfur.
[11:34 AM] rmoretti: @HuubR That's a bit what the "Bad Groups" objective is about. Each round we take a look at the compounds which are coming back, and there's frequently something which our medicinal chemist collaborators will highlight as something which Foldit players are producing which a medicinal chemist would never even consider (because it's just not a feasible compound to synthesize).
[11:34 AM] alcor29: Have any of the results so far been of any use to you?
[11:34 AM] rmoretti: @cjddig: (he/him) I haven't dug into the code for those in that detail, but think it's a little bit of h.
[11:37 AM] rmoretti: The results so far have been excellent for showing what we've not been properly considering, and how to make the tools and puzzle setup better.
As far as coming up with compounds to test in the lab, well, to be honest the useful compounds have been relatively sparse. There's been some "oh, that's interesting" results, but not necessarily a large number of "design of the month" type results.
That's okay. We're learning, and hopefully we can adjust things to get better.
[11:39 AM] HuubR: So our (no, let me rephrase:) my ignorance actually helps you improve your objectives? :-)
[11:40 AM] Susume: (she/her): let's call it 'out of the box thinking'
[11:40 AM] HuubR: :-)
[11:40 AM] alcor29: lol
[11:41 AM] cjddig: (he/him): Will there be other ligand design puzzles besides VHL?
[11:43 AM] rmoretti: VHL is the main target our collaborators are interested in the moment. There's a bit of a push to see if we can get a number of testable compounds, at which point we'll send them out to be synthesized and tested. But once we're done with this initial phase of VHL puzzles, we will likely move on to other targets. (Which ones still to be determined.)
[11:43 AM] alcor29: How is the new UI being accepted?
[11:44 AM] HuubR: I think it is a very good idea to migrate to a single interface, but I also think that "migrate" is an important word here. Judging from the present reactions of long time players, who (understandably) prefer the Original Interface, I feel that the present form has (too) much resemblance to the Selection Interface, and therefore is too far away from the Original.
[11:45 AM] rmoretti: It is true that we leaned heavily on the Selection Interface for the new UI. The reason for that was that the Selection Interface was the preferred interface for the internal Foldit team, and because our understanding is that many of the top Foldit players preferred the Selection Interface (though not universally).
[11:46 AM] cjddig: (he/him): The new UI shortcut is more intuitive (Y for mutate but that's okay) but not familiar. For example, C was for cookbook and X was for cut; now, C is for cut.
[11:47 AM] alcor29: The move tool from selection is kind of awkward. you have to pick it up first from the boxes menu and then locate it. Why can't we keep the original UI way in which we just right click on the exact fulcrum that we want and it just pops up?
[11:47 AM] pc: I use the original interface to move the ligand
[11:47 AM] rmoretti: If there are particular issues where you think the new UI is worse than the Original Interface (aside from just being different), please let us know. We don't consider the new interface to be "finished" by any means, so making it better is something we're interested in.
[11:48 AM] HuubR: Would it be an idea to move the new menu from the left hand side to the tom, where the menu tabs are in the Original Interface, and make the order of the buttons correspond to the Original, as far as possible? That way you would make it easier for a lot of players, who rely on "muscle memory" (I have taken this term from Angus's comment)
[11:48 AM] HuubR: https://fold.it/portal/node/2012337#comment-45514
Developer Preview Release Soon
Developer Preview Release Soon
[11:49 AM] cjddig: (he/him): Actually now I got used to Selection Interface and do like it, so overall, new interface is good.
[11:49 AM] pc: For now we have something good ^^. What we need is more some helps ingame for players that doesn't know chemistry very much
[11:49 AM] HuubR: Another proposal I would like to make: put the action buttons along the left hand side, down from the Foldit logo / main menu button. And give three options: show all buttons, show only those that apply in the present context / selection, and thirdly show only the one that is currently running (the "stop spinner"?). I would prefer the third option, because I use the hotkeys, not the buttons.
[11:50 AM] HuubR: That would place the spinner back in the upper left corner, where I think it belongs.
[11:50 AM] cjddig: (he/him): Agree with pc. New players would need some science backgrounds.
[11:51 AM] HuubR: And please show me how many segments I have selected!
[11:52 AM] pc: Maybe for a biochemist it is easy to know where add an nitrogen to improve clogp, but for me, I just randomly try everything.
[11:52 AM] cjddig: (he/him): Some users just play this for fun (like me)… but now I feel like I should know chemistry.
[11:53 AM] Susume: (she/her): I like the idea of letting us pick which buttons to show (all/some/current) - like HuubR I use mostly hotkeys, so buttons are lost real estate - but for many players they are very useful
[11:53 AM] Nicm25: (improve the glossary. to atoms.)
[11:54 AM] cjddig: (he/him): Well, by the way, where is rebuild? I haven't seen the button since the interface update.
[11:54 AM] HuubR: I learned the use of hotkeys from your YouTube video, @Susume :-)
[11:55 AM] pc: Is orange in "sphere view" means that my ligand is too much hydrophobic ?
[11:57 AM] HuubR: I think that is just the colour for Carbon in the ligand, as long as you have not selected a different colour in the View Ligand menu.
[11:58 AM] rmoretti: The coloring should be controlled separately from the display size (though I admit I don't use sphere view much). If you're using Score/Hydro+CPK coloring, I think the orange is more indicative of the score than being hydrophobic. – Also, if you're using one of the ligand coloring, it might just mean that you have carbons colored orange.
[12:01 PM] rmoretti: Any final questions before we wrap up the office hour?
[12:01 PM] pc: thanks. We use sphere view to visualise shape complementarity. Just see that there is " Ligand Specific" color ^^ (it is more blue)
[12:02 PM] alcor29: Buon lavoro moretti. Thanks.
[12:02 PM] HuubR: Do you have any idea how long it is going to take to find the "Ideal Interface"? :-P
[12:02 PM] rmoretti: The "Ligand Specific" color can be changed from the View Ligand panel (as HuubR mentioned).
[12:03 PM] rmoretti: @HuubR "Ideal" is somewhat subjective. We're approaching this as a continuing process. (And only having the single interface to work on makes it much easier for us to implement those improvements.)
[12:06 PM] Vincera: IMAGE: http://fold.it/portal/files/chatimg/irc_142870_1637427980.png
[12:06 PM] HuubR: What I mean is, when can we expect the Original Interface to disappear in the Main release? In my mind, we're still a long way away from that.
[12:06 PM] Vincera: oops…
[12:06 PM] Vincera: unintended
[12:07 PM] pc: @rmoretti sometimes the visualization of tripple or single bond looks the same.
[12:08 PM] Vincera: Will we lose all of our recipes in the new release?
[12:09 PM] cjddig: (he/him): When will the new webpage come out? I've not heard about it recently.
[12:09 PM] rmoretti: @HuubR The current plan is to promote the current devprev to main once it's finished it's shakedown in devprev, likely sometime next week. (It's following the standard devprev->main migration process.)
[12:09 PM] HuubR: :-P
[12:10 PM] pc: I like the original interface ^^
[12:10 PM] HuubR: (that was my reaction to @cjddig:)
[12:11 PM] rmoretti: @Vincera There shouldn't be any issues with recipes with this release. Where the cookbook is might move around a bit, all your recipes should stay intact.
[12:11 PM] Vincera: D'accord.
[12:11 PM] Nicm25: is big deal, I need to teach others selection interface for migration…!
[12:12 PM] rmoretti: @cjddig: (he/him) The webpage update is proceeding apace. I'm not entirely sure what the timeline is, but we're getting closer.
[12:14 PM] HuubR: I have not tried the Intro Puzzles recently, but I saw that they have been adapted to the new interface. But maybe all those old time players are not very eager to go through all the Intro Puzzles?
[12:15 PM] rmoretti: @Nicm25 The tutorials (Campaign, Education puzzles, Dojo) have been updated for the new interface, and the hope is that they should function as a good introduction to things for people learning the interface. If you have suggestions on how to make them better (or new tutorials which teach things we may have missed), please let us know.
[12:16 PM] rmoretti: Thanks everyone for coming to the Office Hour. It was a pleasure chatting with you.

[agcohn821]

8:57 AM] beta_helix: Hello everyone! Seth Cooper and I will start the last Office Hour of 2021 in a few minutes, but I wanted to share some reading material while you wait. Today's newspost previewing a new Foldit tool: https://fold.it/portal/node/2012556
Sneak Peak - The Trim Tool
Sneak Peak - The Trim Tool
[9:00 AM] Nicm25: hello!
[9:00 AM] beta_helix: Hello Hello!!
[9:02 AM] sethcooper: hello!
[9:02 AM] Josh: (he/him): hola
[9:03 AM] alcor29: Good morning.
[9:03 AM] beta_helix: Let's get the party started… does anyone have any questions they want to start off with?
[9:05 AM] beta_helix: (otherwise I'm happy to start with a question for Seth
[9:05 AM] Josh: (he/him): Okay, so a train leaves Berlin at 230 mph…
[9:05 AM] Nicm25: I'm interested about in the news article is whether it needs to be contiguous segment when trying out new interface WIP. if needs to be continuous, then we may have to think of it as domain proteins.
[9:05 AM] Josh: (he/him): Maybe it's too early for an office hour, west coast is still asleep
[9:06 AM] Josh: (he/him): I think it can be any selection, doesn't have to be contiguous
[9:06 AM] beta_helix: Nicm25 and alcor29 are awake
[9:06 AM] beta_helix: That is correct, it doesn't have to be contiguous.
[9:06 AM] Nicm25: it would be amazing if could be made into separate, detailed selection.
[9:07 AM] Josh: (he/him): I think — but not sure — if you select separate pieces, they will be joined by an open cutpoint
[9:07 AM] Josh: (he/him): or a constraint of some kind to show they are not connected
[9:08 AM] beta_helix: In the blogpost it selected a string of residues, but it also works with a sphere selection (for electron density, for example)
[9:09 AM] beta_helix: Question for Seth, can you tell us more about the new grid tool?
[9:09 AM] sethcooper: okay!
[9:10 AM] sethcooper: it essentially renders a grid along the x-y directions
[9:10 AM] sethcooper: with 1 angstrom cells
[9:10 AM] sethcooper: it centers itself where the camera's focus point is, so it's possible to move it around
[9:11 AM] sethcooper: e.g. if you put the mouse over a residue and press q the grid will move to that residue
[9:11 AM] sethcooper: maybe it could be useful for measuring distances or getting an idea of the scale of things
[9:12 AM] beta_helix: that might be particularly useful if we're able to load in much larger proteins in the future (with the new Trim tool)
[9:12 AM] sethcooper: it might also be more helpful if it's always perpendicular to the camera
[9:12 AM] sethcooper: instead of always x-y
[9:13 AM] beta_helix: Where can we find this new function, Seth?
[9:13 AM] sethcooper: in the view options!
[9:13 AM] Nicm25: "Show grid"
[9:14 AM] alcor29: Currently the grid moves with the protein. MIght be more useful it it was independent? (perpendicular)
[9:15 AM] alcor29: Translation is independent now.
[9:15 AM] beta_helix: I bet Seth is fixing that right now
[9:15 AM] alcor29: lol
[9:16 AM] sethcooper: i was looking at rotation of the grid
[9:16 AM] Nicm25: maybe … : if fix it against the screen, can know it without changing conformation, just change camera angle orientation. (make it parallel to target between 2 points we want to know about)
[9:16 AM] sethcooper: (you do have to select "more view options")
[9:17 AM] jeff101: any news about Foldit's new web page or new tools for Electron Density puzzles?
[9:17 AM] beta_helix: Yes! jeff101!
[9:17 AM] alcor29: A problem we have in judging distances is that they appear different when we zoom around.
[9:17 AM] beta_helix: https://fold.it/portal/node/2012556
Sneak Peak - The Trim Tool
Sneak Peak - The Trim Tool
[9:18 AM] sethcooper: does the grid move with the protein now? it seems like when I move the protein the grid stays where it is
[9:18 AM] beta_helix: the Trim Tool was designed with ED in mind… but should be helpful for all puzzles
[9:18 AM] sethcooper: i mean, it rotates with the protein but not translates
[9:18 AM] sethcooper: i was looking in to fixing the rotation
[9:18 AM] robgee: +1 for the grid, thats handy.
[9:19 AM] alcor29: I like fixing rotation too. But others might have different preferences.
[9:19 AM] beta_helix: I don't want to give a specific date, but I believe the new website should be testable by the first week in January (Josh & Seth, does that sound right?)
[9:22 AM] sethcooper: partly i'd also prefer to not have a lot of extra grid view options since there are a lot already
[9:24 AM] beta_helix: and putting more options in the option.txt file isn't really a solution
[9:24 AM] Josh: (he/him): Sorry beta_helix, I have no new information on the website. Maybe @jflat06 can say more when he wakes up
[9:24 AM] beta_helix: I was just trying to remember from the last Foldit meeting
[9:25 AM] Josh: (he/him): I like that you think I pay attention in those
[9:26 AM] Susume: (she/her): I'd like to be able to fix the grid in space when there is a density, so I can see how I am rotating things relative to it (or don't use the grid for that, and add a set of 3 axes as a separate thing)
[9:26 AM] alcor29: Ditto.
[9:26 AM] jeff101: can we attach bands to points on the grid?
[9:27 AM] sethcooper: not currently but that's an interesting suggestion
[9:27 AM] Susume: (she/her): what a cool idea jeff101
[9:27 AM] beta_helix: then we can create lattices! ;-P
[9:28 AM] sethcooper: the grid isn't really full-fledged "object" in the game at this point
[9:28 AM] sethcooper: i guess it could be used a guide to place bands to space though
[9:29 AM] Nicm25: you mean, something that automatically implements z direction alignment to grid plane when the band is placed?
[9:31 AM] alcor29: A 3D fixed lattice?
[9:31 AM] sethcooper: i think it would be straightforward to add another checkbox for maybe grid fixed or moving
[9:31 AM] beta_helix: that's how protein structure prediction used to be done!
[9:34 AM] jeff101: are any new Foldit designs working in the lab?
[9:34 AM] beta_helix: @Susume: (she/her) and jeff101, did you see the new blogpost?
[9:34 AM] sethcooper: not sure if I can share a video
[9:35 AM] Susume: (she/her): the trim tool, yes I just read it - looks cool!
[9:35 AM] jeff101: I looked at the post about the Trim tool
[9:36 AM] Susume: (she/her): pls tell us more about it
[9:36 AM] jeff101: is there another post you want us to see?
[9:36 AM] sethcooper:
[9:37 AM] Susume: (she/her): cool, seth! that looks very different from the rotating grid
[9:39 AM] beta_helix: (I told you Seth was working on it during the chat )
[9:39 AM] Susume: (she/her): when it's stationary my brain says 'this is for measuring', and when it rotates my brain says 'this is for knowing where you are in space' - feels like two different tools at the moment, maybe at some point they will merge in my brain
[9:39 AM] beta_helix: that was the one I was referring to.
[9:40 AM] beta_helix: The new Trim Tool is work from our newest dev, apetrides, and came from our desire to make Electron Density puzzles easier to work with. Not only will this allow us to post large cryo-EM structures, but it should be useful in all Foldit puzzles (just like cutpoints)
[9:40 AM] Josh: (he/him): @beta_helix just a reminder, folks on the Foldit chat can't see Discord reply metadata, they just get the message.
[9:41 AM] Josh: (he/him): @jeff101 the blogpost beta_helix posted is the one he was referring to
[9:41 AM] beta_helix: thanks!
[9:41 AM] beta_helix: https://fold.it/portal/node/2012556
Sneak Peak - The Trim Tool
Sneak Peak - The Trim Tool
[9:43 AM] beta_helix: apetrides is just putting the finishing touches on it (we want to iron out as many bugs as we can find before letting you find more ) so you should be able to play around with the new Trim Tool soon in devprev (on a specific puzzle)
[9:43 AM] Nicm25: umm, I'm curious about behavior of recipes regarding hidden segments when using WIP, future scripting for ED puzzles.
[9:44 AM] Josh: (he/him): good question!
[9:44 AM] beta_helix: Yes… this is one element that apetrides is working on. Currently, saving/loading doesn't work after trimming.
[9:45 AM] Susume: (she/her): what is WIP?
[9:45 AM] Nicm25: I saw new trimming tool named 'WIP'.
[9:46 AM] beta_helix: Work in Progress
[9:46 AM] Susume: (she/her): oh ok, "work in progress", probably not the final name
[9:46 AM] beta_helix: Why Ih8 Proline
[9:46 AM] Susume: (she/her): lol
[9:47 AM] beta_helix: if beta helices weren't so cool, I probably would have picked an anti-proline username
[9:49 AM] beta_helix: So hopefully at the start of the new year we'll have the grid, this new Trim WIP Tool, and a new website for you to play with . Does anyone have any other questions that we can answer?
[9:51 AM] Nicm25: maybe we need tutorial as well, we'll look forward to it next month, but will there be more campaigns?
[9:51 AM] alcor29: Will the website already include recipe storage? Will cookbook be sortable and searchable?
[9:53 AM] Josh: (he/him): @Nicm25 I can't say much more right now, but next year there will be an entirely new Campaign and lots of new tutorials!
[9:53 AM] Josh: (he/him): @alcor29 I don't know about the cookbook GUI itself, but the website will have some new features for searching and sorting recipes, I think.
[9:54 AM] beta_helix: yes, it's designed to be much better than the current drupal site!
[9:54 AM] Nicm25: (rep) no problem, I just wanted to make an appointment for that text translation beforehand, thanks for answering.
[9:54 AM] alcor29: And will all.macro be stored on the web site?
[9:55 AM] Josh: (he/him): @alcor29 yes, I believe in the new website your recipe library will be stored on our servers and you can browse it on the website
[9:55 AM] Josh: (he/him): And all clients will sync with our servers
[9:55 AM] Josh: (he/him): So you don't have to worry about recipes being local to your machine any more
[9:56 AM] alcor29: Great.
[9:56 AM] jeff101: are Foldit's sponsors still enthusiastic about the project?
[9:56 AM] alcor29: Thanks seth, beta.
[9:58 AM] beta_helix: @jeff101 Yes! apetrides's work is funded by an NSF grant aimed at making it easier to fold into Electron Density… for players and scientists alike.
[9:59 AM] beta_helix: apetrides started in August, and our report is due next summer, so I am sure they will be very enthusiastic about the work so far (it's a 3 year grant)
[9:59 AM] jeff101: that's good.
[9:59 AM] beta_helix: Thank you all!
[10:00 AM] Nicm25: thanks all!
[10:00 AM] jeff101: yes,, thanks for having this office hour
[10:00 AM] Susume: (she/her): @beta_helix there was a recent article about photosynthesis proteins that listed several PDB structures with different kinds of problems - maybe we can help fix some of those
[10:00 AM] beta_helix: that is totally the plan, Susume:! (AJ is working on this too, as you know)
[10:00 AM] Susume: (she/her): yup, looking forward to it
[10:01 AM] beta_helix: Happy end of the year to you all, and we'll see you soon. Keep up the great folding!
[10:02 AM] sethcooper: thanks for the questions and comments everyone!
[10:02 AM] jeff101: :) Happy Holidays!
[10:02 AM] alcor29: Happy New Year all.
[10:02 AM] Susume: (she/her): thanks seth, beta, Josh!

[agcohn821]

[1:30 PM] horowsah: Hi all- here for my office hour!
[1:31 PM] horowsah: I'm a general biochemist, so I'm good for lots of biochem questions
[1:31 PM] horowsah: That said, I'm not a protein design specialist, so can't guarantee my answers there
[1:31 PM] horowsah: I tend to work on ED puzzles and education mode mostly in Foldit
[1:31 PM] spvincent: Hi horowsah
[1:32 PM] horowsah: but in non-things, my lab works on how nucleic acids affect protein folding
[1:32 PM] HuubR: Hi @horowsah :-)
[1:33 PM] horowsah: howdy howdy
[1:33 PM] spvincent: It may be a bit early, but do you know if any of the recent ligand design puzzle solutions showed any promise?
[1:34 PM] horowsah: I don't know- haven't heard anything specific yet. I tend to get a bit of advanced warning but not always
[1:34 PM] HuubR: Do you know if there will be another Electron Density puzzle any time soon?
[1:35 PM] horowsah: hmm, it has been a gap hasn't it
[1:35 PM] horowsah: we've been saving up giving some to use with the Trim tool
[1:35 PM] horowsah: but we probably should do one soon since it has been awhile
[1:36 PM] HuubR: Last one I can find was Puzzle 2007, July 2021 :-o
[1:36 PM] horowsah: Yeah, def time to start getting another one in the works
[1:39 PM] HuubR: Is the Trim Tool ready for testing by us, regular folks? Or is that still too early?
[1:39 PM] horowsah: Last I saw, it had an issue still with disulfide bonds not working right, but in most other cases it seemed pretty much ready for a first iteration release.
[1:40 PM] horowsah: Ultimately, there will likely be a few releases of it where it gets enhanced
[1:40 PM] horowsah: but I would think we're close on the initial version, which should already be rather helpful
[1:40 PM] HuubR: Sounds interesting :-)
[1:41 PM] spvincent: Do any solutions to the HNet puzzles ever get synthesized?
[1:41 PM] horowsah: Sorry spvincent, I actually know very little about that process
[1:42 PM] spvincent: ok
[1:43 PM] HuubR: When you say "how nucleic acids affect protein folding", do you mean folding while the protein is still being assembled?
[1:43 PM] horowsah: yep
[1:44 PM] spvincent: Do you see any kind of role for in yourprotein/nucleic acid interaction work?
[1:44 PM] horowsah: Proteins are made as a linear chain of amino acids, and the folding process sometimes takes awhile so there's lots of opportunities for it to go wrong and so it needs help to get to the right fold
[1:45 PM] horowsah: I very much want to see the nucleic acid work in come to fruition, but first we need to improve how nucleic acids are handled in the game. Unfortunately, that might take quite some time.
[1:45 PM] horowsah: I should say some proteins fold on their own perfectly fine, but some need help
[1:46 PM] jeff101: How do they handle nucleic acids in eteRNA ?
[1:47 PM] horowsah: since they're 2d in eteRNA, they can use a much simpler algortihm than we use to handle 3D atomic-level info in Foldit. A lot of what we need is already embedded in the underlying code, but it hasn't really been unlocked or implemented yet.
[1:47 PM] jeff101: can we build nucleir acids with the ligand design tools?
[1:48 PM] horowsah: jeff101- that would certainly be fun to try, but I doubt it would work well. We can build nucleic acids in right now, but they are only so-so representations, and we lack tools to make them practical for players to manipulate
[1:49 PM] jeff101: is eteRNA useful even though it folds in 2d?
[1:50 PM] horowsah: so 2d representations of RNA are still really useful
[1:50 PM] horowsah: Often that's enough to be able to tell where an RNA is likely doing it's important stuff
[1:50 PM] jeff101: how? why?
[1:50 PM] horowsah: So in RNA, much more function can be interpreted from the secondary structure than in proteins
[1:51 PM] Nicm25: I never imagined future dealing with folding of hybrid proteins with RNA!
[1:51 PM] horowsah: A lot of that has to do with the way evolution has to keep RNA secondary structure intact for it to work
[1:52 PM] horowsah: Knowing the secondary structure of a protein is useful, but it tells you almost nothing about its function by itself
[1:52 PM] horowsah: With RNA, secondary structure tends to be more unique
[1:52 PM] Susume (she/her): (she/her): is there a good reference somewhere for how long proteins take to fold (by size, or by type, or whatever affects it)?
[1:52 PM] horowsah: Still, if you want to, say, design a drug that hits RNA, you still need 3D
[1:53 PM] HuubR: Forgive my ignorance, but is RNA a double helix, like DNA?
[1:53 PM] horowsah: @Susume (she/her): (she/her) I'm pretty sure this has been studied in detail, but I don't have a good reference off the top of my head. In general, multi-domain proteins fold much slower than single domain proteins. Even some single domain proteins, like some GFP variants, can take days to fold outside the cell.
[1:54 PM] horowsah: So RNA and DNA h form double helices, but RNA often forms much more complicated structures too
[1:54 PM] horowsah: They can even form knots
[1:54 PM] HuubR: :-o
[1:55 PM] HuubR: I've been wondering, if you have a helix and a sheet, what would work better for folding: helix at the N terminus, or sheet? In other words, does it help to have the center bit (that would be the helix) first, and then add the beta strands around it?
[1:55 PM] horowsah: I like to think of RNA as having double helix portions (the boring parts) and then parts that are constantly changing, and some that form crazy weird structures
[1:55 PM] jeff101: what are some common RNA 3d motifs (like Greek Key or 4-helix bundles or b-barrels in proteins)
[1:56 PM] horowsah: @HuubR That's a good question I don't know the answer to. Statistically, I assume that the N-terminus would favor one or the other, but I don't know which
[1:56 PM] horowsah: I bet someone out there has checked that out, not sure who
[1:56 PM] Josh (he/him): (he/him): HuubR, regarding RNA, you can maybe see from this link: RNA as two strands (double helix) doesn't need to always bond with its adjacent partner, which can make for some interesting bubbles and loops and things
[1:56 PM] Josh (he/him): (he/him):
[1:56 PM] horowsah: @jeff101 the basic building block in RNA is often called a stem-loop
[1:57 PM] Susume (she/her): (she/her): is the explosion in cryo-EM giving you access to more RNA-protein complex structures than what x-ray crystallography can do?
[1:57 PM] horowsah: …but there's less definitive classification of RNA structures than for proteins
[1:57 PM] jeff101: is there a database for RNA similar to the SCOP database for proteins?
[1:58 PM] horowsah: @Susume (she/her): (she/her) I think so. It took awhile for cryo-em to catch up on RNA, and there are still some technical kinks being worked out, but pretty soon I expect it to help in a big way.
[1:59 PM] horowsah: @jeff101 I don't think there is a direct comparison for the SCOP database with RNA. I think in part it's that when we usually look at RNA structures, we look at small pieces of them at a time. It's not as easily modular as proteins are, but we have to study them that way because it's what we can do. In total, our knowledge of what whole RNAs look like when put together is still lacking
[2:00 PM] horowsah: I find how proteins and RNA bind each other fascinating, because it's typically a lot more loosey-goosey than how proteins bind each other
[2:00 PM] spvincent: eterna has some lab puzzles that aim to change the ribosome so as to enable it to synthesize things other than standard proteins. Concentrating on RNA naturally enough. But what about all the RNA/protein interactions? Will the ribosomal proteins have to be changed as well?
[2:00 PM] jeff101: how has DeepMind / AlphaFold affected your research?
[2:01 PM] horowsah: @spvincent In the ribosome, the RNA is usually more important, but I would think some engineering of h would help
[2:02 PM] horowsah: @jeff101 so far not in a huge way, but I expect that to change with some upcoming crystallography projects. With the RNA-protein interactions work it probably will soon as we start to try and come up with starting models of how they bind
[2:03 PM] HuubR: Does RNA always fold up the same way? (thanks for the interesting picture, Josh :-) If not, would that be a problem for Cryo-EM?
[2:03 PM] jeff101: like tRNA .. is it a pretty rigid structure?
[2:04 PM] horowsah: RNA and proteins h typically fold up the same way, but there are classes of h that don't and are intrinsically disordered. Some RNAs are quite rigid, but some switch between several different structures, which is a bit more common in RNA than in proteins. Those RNAs are sometimes very important biologically and are quite hard to study as a result of their switching
[2:06 PM] jeff101: is there such a thing as membrane RNA, where the hydrophobics are on the outside?
[2:07 PM] horowsah: That would be cool! Sadly, I've never heard of it before, and I think I would if it had been discovered. Some prokaryotes tether DNA to the membrane, though.
[2:07 PM] jeff101: do RNA complexes have a hydrophobic core and a hydrophilic exterior?
[2:08 PM] horowsah: Typically, the binding of RNA and proteins is mostly electrostatic (hydrophilic) in nature. Still, that's an overgeneralization, and often you see key hydrophobic contacts.
[2:09 PM] horowsah: There are proteins that can unwind RNA, and then sometimes you see the bases from the RNA stacking with aromatic hydrophobics from proteins
[2:10 PM] Susume (she/her): (she/her): @jeff101, there was a recent paper where they found that nucleic acids on the cell surface could be decorated with glycans (sugars), just like proteins and lipids can be decorated with glycans - so some RNAs do get expressed on the outside of cells (the other meaning of transmembrane - they cross the membrane, rather than sticking in the membrane)
[2:10 PM] jeff101: what about intracellular membranes?
[2:11 PM] jeff101: with only 4-5 RNA/DNA bluidling blocks, how do they make such diverse structures? are certain A T G C U considered hydrophilic or hydrophobic?
[2:12 PM] horowsah: So protein structural diversity is definitely greater than RNA. The thing that RNA has going for it here, though, is greater degrees of freedom within it. Basically, a rama map of an RNA shows it to be more flexible in the number of arrangements it can take compared with a protein of the same size
[2:13 PM] horowsah: on the intracellular membrane question, i'm not sure, I'd have to do some reading there
[2:15 PM] Nicm25: I've heard that we can create an arbitrary shape by creating like zipper with pair matching combinations of RNA bases… 'DNA origami'
[2:15 PM] horowsah: Yes, basically. DNA Origami
[2:16 PM] horowsah: it's pretty darn cool looking
[2:16 PM] spvincent: Wasn't there a game called Nanocrafter that tried to do something similar?
[2:17 PM] horowsah: yeah… not sure how it worked though
[2:17 PM] spvincent: I'm not sure if it still exists
[2:18 PM] HuubR: https://citizensciencegames.com/games/nanocrafter/
Citizen Science Games
CSGamer
Play Nanocrafter, use pieces of DNA to build everything !
Play Nanocrafer, a citizen science game, use pieces of DNA to build everything from computer circuits to nanoscale machine.
Image
[2:19 PM] Josh (he/him): (he/him): scooper made that, I think! I doubt it's still playable
[2:20 PM] spvincent: there's a message on their website saying the project has ended
[2:22 PM] horowsah: yeah, that makes sense
[2:26 PM] Nicm25: in the future, I would like to see ability to load only some residues from solution sharing in ED puzzles to improve group work efficiency.
[2:26 PM] horowsah: That would be nice… we can put that on the list.
[2:27 PM] HuubR: Distributed Gaming? :-)
[2:27 PM] horowsah: we've talked about that a bit with in the past, but not sure if it really will happen
[2:28 PM] jeff101: I guess it'd be like the Trim Tool, but mutiple people can grab certain parts of the whle protein and work on them separately.
[2:28 PM] Nicm25: I came up with that idea because I didn't think huge protein could be processed by single computer
[2:28 PM] horowsah: yeah, pretty similar
[2:28 PM] jeff101: Later, soething puts all the parts back together.
[2:29 PM] jeff101: If several people work on the same piece, how to pick which persons' piece to put back into the whole protein?
[2:30 PM] horowsah: yeah that would be an interesting thing to figure out how that would work
[2:30 PM] horowsah: I have to go, thanks for the great chat all!
[2:30 PM] HuubR: Thank you, @horowsah!
[2:31 PM] Nicm25: thank you all.

[agcohn821]

[9:01 AM] sciren: (he/him): Hello Everyone! I hope you all are having a good day/night. It is a bit colder here than I'm used to, but I've got coffee to help
[9:01 AM] Dudit: (he/him): Hi @Sciren
[9:02 AM] HuubR: Good morning @Sciren
[9:02 AM] jeff101: Hi @Sciren
[9:02 AM] Susume: (she/her): hi @Sciren!
[9:03 AM] jeff101: @Sciren what are you working on these days?
[9:03 AM] Susume: (she/her): pc and Huubr I love the idea of testing whether wiggle allows the ligand to enter the pocket when the opening is tight - @Sciren, any comment on whether this experiment is real-world relevant?
[9:04 AM] sciren: (he/him): @jeff101 Currently I'm working on UI enchancments/bug fixes, along with adding improved features to help with the next round of Small Molecule Design Puzzles.
[9:05 AM] sciren: (he/him): @Susume I just want to make sure I understand. You want to leave the ligand outside the pocket and use wiggle to see if it will settle/dock into the pocket?
[9:06 AM] Dudit: (he/him): @Sciren I think the time counter / time left until the puzzle expired is missing in the new Foldit user interface ( it's now hard to know when to submit to the scientist the last moment before the puzzle expired )
[9:07 AM] Susume: (she/her): yes, pc was using it with a rubber band to see if the small molecule could get inside, since the opening to the pocket in his protein is partly covered
[9:07 AM] sciren: (he/him): @Dudit The timer should appear when 24 hours are left on the puzzle. Is this happening for you?
[9:08 AM] pc: Hi. Yes we are interested at what happen in real life when ligand entering, to help us create better designs that work in lab ^^
[9:09 AM] pc: So we try to do a pseudo physic simulation ^^
[9:09 AM] sciren: (he/him): Interesting @Susume I will have to check to see how viable this would be. I don't believe wiggle is set up with this behavior in mind. There are other ways to make this happen in Rosetta, but I really like the ingenuity of forcing docking with the tools that you have. This is something I can look into further!
[9:10 AM] Dudit: (he/him): @Sciren Thank you, the timer is appeared now
[9:11 AM] sciren: (he/him): No worries! @Dudit I was ready to jot it down as a bug just in case.
[9:12 AM] sciren: (he/him): @Susume this also brings up a good point, protein-ligand docking tools could be useful down the road.
[9:12 AM] pc: the new "neural net objective" is interesting. Do you plan to add score on it ?
[9:13 AM] HuubR: :-)
[9:13 AM] sciren: (he/him): @pc It is indeed really interesting! Unfortunately, I'm not sure about the progression plan for the neural net objective, but it is something I can ask about and let you know.
[9:15 AM] pc: Maybe scientist doesnt want we use too much AlphaFold for points. But a new AlphaFold request can be needed for points only if there is a mutation for exemple.
[9:15 AM] pc: (if there is points in "neural net objective I mean)
[9:16 AM] sciren: (he/him): As a gamer, I really like objective completion, but I also understand there are times when we want these objectives to be inherent in the scoring.
[9:17 AM] pc: I often share to scientist a lower score solution than my top score solution. Because the top score solution have often bad AlphaFold
[9:22 AM] pc: For binder puzzles, if I understand, scientists are mostly interested by interface(from player shared protein), and create new proteins with same interface (from library). Now with AlphaFold, maybe it is interesting to get more variants from players with moslty good AlphaFold similarity (and correct interface) ? (and not only the best protein with best interface)
[9:24 AM] Dudit: (he/him): @Sciren: I think maybe there is a way to get an aggregate scoring system that is combining Rosetta & AlphaFold score into one whole score
[9:24 AM] pc: on the 32000 tested proteins , maybe we can add more proteins from players, to experiment how AlphaFold help ?
[9:24 AM] sciren: (he/him): @pc I'm not certain which is the best approach for this to be honest, but I it is something that I can pass along to get a bit more information about.
[9:25 AM] pc: thanks
[9:25 AM] HuubR: About bug fixing: I think georg137 had a very interesting suggestion in a feedback item last December (link below), to make a list available of bugs (and possibly suggestions) that have already been reported, and maybe also mentioning which ones are being worked on.
[9:25 AM] HuubR: https://fold.it/portal/node/2012568
[9:26 AM] HuubR: If such a list is accessible on the web site, it would be easier for us, players, to check whether a bug that we might encounter is already known or not, and possibly what its priorities are.
[9:28 AM] sciren: (he/him): @Dudit Before changing/adding to the existing scoring, we would have to make certain that it is the best approach. I'm sure there are pitfalls that we want to avoid before going down that path. But it is something to consider, and something I can bring up to the devs.
[9:28 AM] Dudit: (he/him): @Sciren Thanks
[9:32 AM] pc: A suggestion for AlphaFold score, is to limit score after 80% confidentiality. I think maybe it can be enough.
[9:32 AM] sciren: (he/him): Absolutely, @HuubR feedback loop is a troublesome thing in any user facing project, and finding the best solution is tricky. It is something that we have discussed as a team, and we will do what we can to make sure we are addressing as many concerns as possible.
[9:32 AM] Vincera: May we please have the date/timer displayed during the full length of puzzle? As in blitz chess, eg. If I have 3-4 apparati set-up in workstation, I can walk from one to the other and on sight, see this notation. Should not have to incur further clicks; Multiple puzzles synchronously played with diff endtimes; we need constatnt display in the grey
[9:34 AM] HuubR: I agree with Vincera: what reason is there NOT to display it all the time?
[9:34 AM] Vincera: And some of ud devote device strictly to DEVPREV.
[9:35 AM] pc: When I use recipe with "wiggle + mutate + band" for exemple, score increase, but there is sometimes new buns after. Is there a solution or we can add a new slider "BUNS importance" ?
[9:35 AM] Vincera: Further, the time left when competing vs player in diff time zone will add to decision-making.
[9:37 AM] sciren: (he/him): @Vincera I will definitely note your suggestion. Understanding the broader use cases is extremely helpful in making these design choices.
[9:38 AM] Vincera: If I am up at 05h30 in my timezone to check the European stock market intime to make my trade for USA open, I can save this brain power for other part of the day/week. Thanx.
[9:39 AM] sciren: (he/him): @HuubR To answer your question, one of the design choices that was made was to attempt to declutter the existing Score Board which included the existing in puzzle timer. That being said, there should be a way to add this back in as an optional UI element for players who want to see it always displayed.
[9:41 AM] HuubR: Thanks. That brings up an interesting point: why do we have to set several options time and again when we start a client?
[9:43 AM] jeff101: Here is an old scoring suggestion: instead of counting non-ideal loops, count segments that are in non-ideal loops. Right now, there can be a non-ideal loop covering segments 11-30, and it gets less penalty than several non-ideal loops covering segments 11-14 18-19 and 27-29
[9:43 AM] HuubR: :-)
[9:43 AM] sciren: (he/him): Ah yes @HuubR . That is actually a bug that I am working on currently with @Josh: (he/him) . We were able to locate an issue that was causing the options to not save correctly. We're hoping to get that out the players as soon as possible.
[9:43 AM] Josh: (he/him): I have been summoned!
[9:43 AM] Josh: (he/him): Oh, the option saving bug, yes.
[9:44 AM] sciren: (he/him): Ha! With the power of your exceptional coding skills @josh (he/him) himself has appeared!
[9:44 AM] Josh: (he/him): mostly with the power of pings
[9:44 AM] Josh: (he/him): currently virtually attending the Boston Festival of Indie Games
[9:45 AM] Josh: (he/him): someone here has nearly the same voice as you sciren:</b> and it threw me for a loop for a second XD
[9:45 AM] jeff101: having the non-ideal loops penalty depend on the number of segments would help recipes gradually fix non-ideal loops
[9:46 AM] HuubR: Thank you, @Sciren and @josh. Looking forward to trying the next DevPrev release :-)
[9:46 AM] sciren: (he/him): @jeff101 I'm not sure about this scoring issue, but it is something that I can pass along to see what can be done.
[9:46 AM] Susume: (she/her): I like jeff101's idea about length of non-ideal loop being part of the penalty
[9:46 AM] sciren: (he/him): Alas, I am not in Boston at the moment. Though I would like to go some day!
[9:46 AM] jeff101: it seems to me that a non-ideal loop covering segments 1-7 should have less penalty than one covering segments 1-14
[9:47 AM] Susume: (she/her): that would also help with the situation where one resiude at the end of the protein is getting hit with ideal loop penalty - if length were taken into account, the penalty for a terminal resiude would at least be smaller
[9:47 AM] jeff101: right now, h count as 1 non-ideal loop, and recipes can't tell the difference just from the Foldit penalties
[9:49 AM] sciren: (he/him): I see what you are saying @Susume and @jeff101. Right now there is no distinction of total non-ideal loop segments.
[9:49 AM] HuubR: I would just hope that a bad loop at the end of the momomer in a Symmetry puzzle would not result in 2n+1 residues counting as bad loop. At the moment, that is the way such a situation is visualized :-o
[9:51 AM] jeff101: @HuubR is the n in 2n+1 residues the length of one monomer?
[9:51 AM] HuubR: Yes
[9:51 AM] Susume: (she/her): re: decluttering the GUI - to me it is only decluttered if I get some real estate back - having an enormous score number that takes up just as much space as the older more-informative panel is not less cluttered imo
[9:53 AM] Susume: (she/her): all the non-movable GUI elements feel very big on my screen, the devs must play at a higher res than I do
[9:54 AM] sciren: (he/him): Thank you @Susume You are correct. It is the same size as the old score panel. In this case the decluttering was in reference to how much information was being pushed into one panel. So the idea was to find the best homes possible for all of this information, because non of it is extraneous.
[9:55 AM] sciren: (he/him): Yes indeed, resolution differences are a challenge, and one that we are always trying to take into account, but unfortunately it doesn't always get adjusted the way we hope.
[9:56 AM] spvincent: Any chance of providing alternatives to the current colouring scheme? I think the saturated pastel look of a fully selected protein is perfectly hideous.
[9:57 AM] Vincera: Ditto that, spvincent. Puke factor.
[9:57 AM] Dudit: (he/him): @Sciren Maybe resolution scaling in % ?
[9:58 AM] sciren: (he/him): I actually have been looking into color schemes a bit lately, and I will add selection as one of the elements to go along with that.
[9:58 AM] Josh: (he/him): it's Virtual this year!
[9:59 AM] jeff101: @josh what is Virtual this year?
[10:00 AM] sciren: (he/him): @Dudit Indeed. That is something that has been discussed, and I think it will make its way in the future. Ideally, as many UI elements as possible would scale accordingly to screen size.
[10:00 AM] HuubR: Can we set the colours of selected residues in the file theme.txt? I see a ton of RGB values there, but it is not obvious what they represent.
[10:00 AM] Dudit: (he/him): @Sciren Thank you
[10:01 AM] sciren: (he/him): @HuubR I'm not sure of the status of the theme.txt file and subsequent support. I believe it needs some updating.
[10:01 AM] HuubR: :-))
[10:01 AM] Josh: (he/him): this has some info: https://foldit.fandom.com/wiki/Theme
Foldit Wiki
Theme
The theme.txt file is used to customize your color scheme within Foldit. The file can be found inside the base Foldit directory (folder). To make a change, edit the values while Foldit is not running, otherwise your changes will be lost when the game exits.
[10:02 AM] Josh: (he/him): oh sorry @jeff101 I was responding to @Sciren that the Boston Festival of Indie Games is virtual
[10:03 AM] spvincent: tx Josh: I'll take a look
[10:06 AM] jeff101: Thanks for having this Office Hour.
[10:06 AM] sciren: (he/him): Alright everyone, unfortunately I am going to have to go. I just wanted to thank all of you for your questions, and also take a moment to thank you for all of your feedback. As I have mentioned before, I am an avid gamer, and I know first hand how it can be to adjust to new user interfaces. Your comments and concerns have been and are extremely helpful in making sure provide you all with the best experience possible. So thank you again. I hope all of you have a wonderful day! Happy Folding!
[10:06 AM] Dudit: (he/him): Thank you so much @Sciren & @josh
[10:07 AM] Susume: (she/her): thanks @Sciren!
[10:07 AM] HuubR: Thank you, @Sciren and @josh, and keep up the good work :-)
[10:07 AM] pc: thanks

[agcohn821]

11:00 AM] bkoep: Hi all, welcome to Office Hours!
[11:00 AM] jmbrownlee333: hi
[11:00 AM] bkoep: I'll be here for the next hour or so to field questions and chat about
[11:01 AM] jmbrownlee333: I'll jump in. I have questions about the latest spate of ED puzzles. Comment on the usefulness of player solutions and also how are player solutions assessed in terms of the diffraction data?
[11:02 AM] bkoep: @jmbrownlee333, certainly! The recent ED Reconstruction puzzles are based on protein structures that are already published, but have some issues
[11:03 AM] jmbrownlee333: Issues? such as
[11:05 AM] bkoep: Usually these issues have to do with poor chemical geometry (unrealistic bond angles, etc.). These kinds of errors can result from over-fitting a model to questionable experimental data. These issues are more common in early crystal structures (pre-1990s, maybe), but they still sometimes crop up if the crystallographer or microscopist isn't careful
[11:07 AM] jmbrownlee333: so then model bias in the phases? The reason i ask is there were instances in 2108 where best-scoring conformations were at odds with the ED, in my solutions and my teammate’s.
[11:08 AM] bkoep: Sometimes it is difficult to fix these issues. Naively correcting the geometry can cause the "model-data agreement" to drop. The idea is that players can effectively reconcile differences between the model quality ( score) and experimental data (ED cloud)
[11:09 AM] bkoep: @JMbrownlee, yes there can be model bias in the phases that cause artifacts in the ED cloud—although we take some steps to limit model bias when we prepare ED puzzles
[11:10 AM] bkoep: in 2108 where best-scoring conformations were at odds with the ED, in my solutions and my teammate’s.
That is interesting! can you say more about this?
[11:11 AM] bkoep: Do you mean that there were parts of the ED cloud that were unoccupied by your best-scoring model?
[11:12 AM]alcor29: I get that mutating 20% of the residues will cause a significant jump in the confidence level. The question to us becomes which 20%. Sometimes NN paints a large portion of the mer as red. The permutations are daunting. Trial and error could work but very time consuming. Could the basic mutate programs read the NN data? Any other suggestions?
[11:12 AM] jmbrownlee333: For instance, lys 7 densiy was clearly heading towords an h-bond with the sulfate ligand. But any shaking or wiggling took the side chain out of the density.
[11:13 AM] jmbrownlee333: typing not my strongest skill ;)
[11:13 AM]alcor29: Sorry. Didn't mean to interrupt.
[11:14 AM] pc: @bkoep: Hi. Sometimes BUNS in a design puzzle, is a bit outside of the contact area.
Do you think that the water molecule is actually ejected ?
Maybe the protein can still bind more easily than if the BUNS was in the middle of the area ?
[11:14 AM] jmbrownlee333: and gln 69 loop had that side chain going in a different direction.
[11:17 AM] bkoep: lys 7 densiy was clearly heading towords an h-bond with the sulfate ligand. But any shaking or wiggling took the side chain out of the density.
Yes this is the type of thing that might cause problems, especially if the LYS requires a strained backbone or sidechain to make the H-bond. I don't know about this specific case, but if Wiggle has trouble it can mean the solution is stuck in a local minimum. For example, maybe an entire loop needs to be remodeled or Remixed so that the LYS backbone is better positioned to make the H-bond. You might have to sacrifice some ED fit in one area in order to make greater gains somewhere else.
[11:20 AM] jmbrownlee333: Last question on this. How will researchers use player solutions?
[11:20 AM] bkoep: @alcor29, yes that is the question! Even if the NN Objective tells you what region is bad, what mutations will fix things? Right now, you are sort of on your own to try different things. If there are problems with your original solution, it might be enlightening to look at the AlphaFold prediction. For example, if the AF prediction has buried a particular residue, you could try mutating that residue to something polar and charged—to "destabilize" the structure predicted by AlphaFold.
[11:22 AM] bkoep: In the longer term, we are working on a deep learning-based Mutate tool that may be more effective at improving AF confidence (possibly at the risk of a score decrease). Stay tuned for more developments in the coming months!
[11:24 AM] zippyc137: I had 83.7% confidence in AF with a design with 2 buns, solved the buns, got a higher score in , uploaded in AF again and got 73.7% confidence, I know buns are bad so this confuses me what score to follow
[11:24 AM] brunokestemont: Hi, I noticed that AF on monomer is'nt always a good indication of what will happen later on with symmetric. In symmetric, i usually mutate parts of the monomer (which gives something else), and the latest good scoring solutions perform better in AF as well.
[11:25 AM]alcor29: I think score decrease will be necessary, but that will apply to all so there would be no problem.
[11:27 AM] bkoep: @pc Yes, measuring BUNS can be a little unstable, due to approximations in our surface area calculations. This means that the BUNS Objective will occasionally flag an atom as buried even if water can technically get in and satisfy the atom. Unfortunately, if we want to remain responsive and (relatively) snappy, we have to make concessions like this.
[11:28 AM] pc: thanks
[11:29 AM] bkoep: I know some designers at the IPD do prioritize BUNS that are deeply buried (or "very buried" - VBUNS) over more marginal BUNS atoms ("surface buried" - SBUNS). I think that distinction may be a little too intensive for , but it is something we can consider in the future.
[11:30 AM] pc: @bkoep: In design puzzles, objective is to have more than 400 CS and less than -40 DDG.
But it is useful to share a solution with 350 cs and -30 ddg (but with no buns) ?
Is it interesting for scientists for other reasons than be selected for lab ?
[11:31 AM] brunokestemont: same question: is it interesting to share to scientists a 0 BUN solution (even with low HBonds) ?
[11:32 AM] bkoep: @JMbrownlee333 After ED Reconstruction puzzles, we do some further analysis with solutions to see how well they agree with the fundamental experimental data (i.e. we run crystallographic refinement with the models and evaluate R factors, Rama outliers, bond geometry, etc.). We hope to see that solutions improve the fit-to-data and geometry issues that were present in the published structure.
[11:33 AM] jmbrownlee333: Thx.
[11:37 AM] bkoep: @zippyc137 Yes, you are right! It is very confusing to get conflicting data from AlphaFold and the score and Objectives, etc. Unfortunately I don't have the answers to resolve your confusion. The team is working on new tools that will make it easier for players to try different things, but ultimately I think it's up to the player try and find something that works. Maybe we'll learn something from you!
[11:39 AM] zippyc137: thx, we"ll keep on experimenting:)
[11:40 AM] bkoep: @brunokestemont Yes, the AlphaFold confidence can be unpredictable. Sometimes a seemingly-innocuous mutation can cause a drastic change in confidence. Other times you can completely redesign your protein with hardly an effect on confidence. Unfortunately, the AlphaFold network is a bit of a mystery (this is an issue with deep neural nets in general), so we can't say exactly why AlphaFold "thinks" the way it does
[11:40 AM] brunokestemont: Now suppose I load an "original" from AF: I get the confidence noticed on screen. If I completely change the 3D structure without mutating anything, AF "score" will remain the same.
[11:41 AM] pc: zippyc137 : what I think (tell me if it is true) : big score in and low score in AF means : score for this specific folding is true, but this folding never appears in reality (or have less chances)
[11:43 AM] brunokestemont: AF rely on published data. Is there a risk of "autoreferencing" and repeating past errors ? (AF would not help us to "invent" original solutions)
[11:45 AM] bkoep: @pc Yes absolutely, please share "imperfect" solutions. If you've pushed a solution as far as it will go, and you just can't reach the Objective targets, then we'd like for you to share the solution and then move on to another attempt. The Objective targets are set at a "high bar" and we expect they will be difficult to satisfy.
[11:47 AM] zippyc137: @pc, I guess so too, so the scoring in should also change accordingly
[11:48 AM] bkoep: @brunokestemont Re: refolding the original - Yes, that's right. You can refold your solution all you like, but if you make no mutations then the AlphaFold prediction will stay the same.
[11:49 AM] alcor29: How have recent results from looked regarding NN / AF compliance vs scores and real foldability? Or is it too early to tell yet re NN usability?
[11:50 AM] pc: Is the AF prediction is precise ? Sometimes I try to create protein that keep a good similarity to have more chance to bind without deforming too much.
But I dont know if AF is enough precise for that.
[11:51 AM] alcor29: " but if you make no mutations then the AlphaFold prediction will stay the same." Doesn't this imply that AF is merely reading the string of AAs?
[11:52 AM] bkoep: @brunokestemont Re: autoreferencing - Good point, when evaluating a trained model (like AF) it is important to consider overfitting to the training data. However, the DeepMind team (and others) have shown pretty convincingly that AlphaFold is useful for extrapolating beyond the training data. I do not think we have to worry about AlphaFold showing bias against invented, original solutions
[11:52 AM] brunokestemont: In his video, nspc suggest to move the monomer away before to submit to AF. I'm not sure it's necessary. What do you advise?
[11:52 AM] Susume (she/her): yes, AF is only reading the string of AAs
[11:52 AM] pc: hum. I dont remember that
[11:53 AM] pc: but maybe I tested something else bruno
[11:53 AM] Susume (she/her): the similarity score reported by looks at your 3D structure - but AF and the confidence score do not
[11:53 AM] brunokestemont: https://www.youtube.com/watch?v=mrCuMu1ay_g
[11:54 AM] brunokestemont: ah ok thanks susume
[11:54 AM] pc: ah yes, it is the first base protein. I try to create a stable one with good AF before bind to target
[11:55 AM] alcor29: @Susume (she/her) Why does that seem a shocking revelation to me? So it's just saying this string is not in my library, Therefore it's not workable. But if that's so does it not imply then that our "creativity" is extremely limited, and that automation will work better?
[11:56 AM] Susume (she/her): it is not really saying 'not in my library' - it takes your string and makes a 3D prediction of how those AAs will fold - even if it has never seen that string before
[11:56 AM] bkoep: @alcor29 We don't have any data yet about how the Neural Net Objective affects wet lab results. It does seem like players overall are using the Neural Net Objective productively, to make solutions with improved AlphaFold confidence
[11:57 AM] alcor29: Re PCs procedure: I've had 85% stand alone and then 66% when adapted to symmetric chains.
[11:57 AM] pc: NN helped me fix some soluitons more easily. But it is hard when all is blue and confidence still at 75%. I have to see prediction for that.
[11:58 AM] pc: and continue fix
[11:58 AM] Susume (she/her): AF has been shown to be quite good at predicting the 3D shape for a novel designed string of AAs - and also of reporting low confidence for strings that are unlikely to fold well
[11:58 AM] alcor29: Ah thanks @Susume (she/her) .
[12:00 PM] bkoep: @pc Yes AlphaFold predicted structures are remarkably precise, and that is the main reason for all of the hype about it. In relation to other DNNs (like trRosetta), AlphaFold is actually no better at predicting design success vs. failure (will the design fold?). But, if the design does fold, then it probably folds just as the AlphaFold prediction says it will. Hopefully that makes sense?
[12:00 PM] ichwilldiesennamen: I've notice with symmetry-puzzles that when you have a good monomer with very good AF-values around the 90s then you can in most cases keep those values when attaching the copies. But you have to be careful not to mutate too much. A bit is alway s necessary (especially to get the HBnet)
[12:01 PM] pc: yes thanks. so my strategy can helps to have more successful binding maybe. I will try again sometimes ^^
[12:02 PM] brunokestemont: Me to I usually first design a very good monomer for bonus, AF and score. Then I try to bind or to make HBonds.
[12:03 PM] bkoep: @brunokestemont @pc It should not matter at all if you move the monomer away from the target before submitting to AlphaFold. Under the hood, completely removes the target before setting up the AlphaFold prediction. In fact, you are noticing that translating/rotating the monomer gives you different results, then we should treat that like a bug.
[12:04 PM] alcor29: We know how to make a monomer with 95%. Trick is to interface it to others. How good is AF at checking assemblages of mers?
[12:04 PM] pc: brunokestemont : yes, want I test with AF when I detach monomer is only create a first stable protein (but ignoring the target)
[12:05 PM]alcor29: Or is the question even applicable?
[12:06 PM] brunokestemont: @pc ok in order to avoid a target interfering with the monomer when trying to design a "perfect" monomer.
[12:07 PM] pc: yes ^^
[12:07 PM] ichwilldiesennamen: Shouldn't the best way to do the interface to the symmCopies be to have an as small as possible interface-area? Like this only very few SCs will have to be mutated (from hydrohpyl to hydrophob). And the AF-vaules will change only mildly.
[12:08 PM] brunokestemont: alcor29 I think that AF only test the monomer part, not helping for the interface
[12:08 PM] bkoep: @alcor29 Re: multimers - Great question! Yes, other researchers have found that AlphaFold is often capable of predicting multimer assemblies—although this is not currently enabled in puzzles. We may be able to add it at some point, but there are some problems we will need to iron out (for example, we would not be able to load AlphaFold multimer predictions into symmetry puzzles—what if the AF prediction is not symmetric?)
[12:10 PM]alcor29: Thanks very much bkoep:. It has been very useful
[12:10 PM] jmbrownlee333: Yes, thanks for giving up part of your Sunday
[12:10 PM] pc: thanks ^^
[12:11 PM]alcor29: @ichwilldiesennamen. Good idea. I will try it.
[12:13 PM] ichwilldiesennamen: @alcor29 Yes give it a try. I previously packed everything together as much aspossible but AF showed that this is not advisable. Now I try to have only 2 SCs as core-elements on the interface and build my HBnet with those. Works pretty well and AF stays up
[12:13 PM] bkoep: @ichwilldiesennamen Yes I like this approach of focusing on small interfaces in symmetric designs. The main difficulty to watch out for is that small interfaces may not have enough specificity. A lack of specificity may allow your protein to form off-target assemblies. Recall the infamous crystal structure of NinjaGreg's symmetric trimer design that folds into a tetramer: https://youtu.be/dxfE3A05rzA?t=190
YouTube
Lab Report #22: Unexpected Lab Results
[12:15 PM] bkoep: Alright I will head out now. Thanks all for the questions and some great discussion about protein design!
[12:15 PM] bkoep: I'm looking forward to your latest designs. Don't forget to Share with Scientists!
[12:16 PM] ichwilldiesennamen: Thx bkoep: for your time!

[agcohn821]

[10:01 AM] beta_helix: Hello everyone! Today's Foldit Office Hour will focus on discussing the new Electron Density Reconstruction puzzles!
[10:02 AM] jeff101: Hi beta_helix!
[10:02 AM] beta_helix: Hi jeff101!
[10:03 AM] beta_helix: How are things?
[10:03 AM] jmbrownlee333: Hi. So how are we doing?
[10:03 AM] beta_helix: Hi jmbrownlee333. All good on this side.
[10:03 AM] beta_helix: What have you thought of the recent ED Reconstruction puzzles?
[10:05 AM] jmbrownlee333: I took particular interest in RNase structure. The density and foldit score seemed to point in different directions in certain instances
[10:05 AM] beta_helix: So what did you do in that case?
[10:06 AM] jmbrownlee333: I sometimes wish for an fo-fc map.
[10:06 AM] beta_helix: Hahahahahhahahaa
[10:06 AM] jmbrownlee333: Since its a game i let score in foldit dictater what i do.
[10:06 AM] beta_helix: Funny you should mention that… as that is stage 2 of the ED grant we got from the NSF
[10:07 AM] jeff101: I follow the score too.
[10:07 AM] jeff101: what do you mean, stage 2?
[10:07 AM] beta_helix: It's tough because the density is supposed to be correct, but we know that there are artifacts
[10:07 AM] beta_helix: (sorry, the grant has different sections. The first part is the trim tool)
[10:08 AM] beta_helix: https://fold.it/portal/node/2012556
Sneak Peak - The Trim Tool
Sneak Peak - The Trim Tool
[10:08 AM] beta_helix: which should be out early next month (maybe in devprev even earlier)
[10:09 AM] jmbrownlee333: So how are we doing as a community. Have we impoved R and Rfree on our player solutions?
[10:09 AM] beta_helix: The other idea in the grant (stage 2) is to be able to allow you to refine your map based on your model
[10:09 AM] beta_helix: YES, you have!!!
[10:10 AM] jmbrownlee333: Is that the metric you are interested in?
[10:11 AM] beta_helix: This is only preliminary (hence no blogpost yet, as we need to run more puzzles) but most of your results have significantly improved over the starting PDB models (using multiple metrics).
[10:12 AM] beta_helix: For more details, these were the metrics we used in the Foldit "Determining crystal structures through
crowdsourcing and coursework" paper, (Figure 2): http://www.cis.umassd.edu/~fkhatib/Papers/NatureComm.pdf
[10:13 AM] jmbrownlee333: How have the puzzles been chosen for the latest ED puzzles?
[10:14 AM] jeff101: have they all used X-Ray Diffraction Data with the same resolution?
[10:14 AM] beta_helix: We've run 4 so far this year.
[10:16 AM] beta_helix: No jeff101, they have had different resolutions. We specifically picked a bunch of PDBs they did not seem great. It wouldn't be as useful as an initial test set to pick perfectly solved structures
[10:16 AM] beta_helix: How have you felt folding from a starting model, as opposed to previous ED puzzles that were most often giving as an extended chain?
[10:17 AM] jmbrownlee333: of course the starting model makes it so much easier.
[10:18 AM] beta_helix: What about the fun factor? Does it take away from the challenge, or on the contrary: it is less frustrating?
[10:19 AM] jeff101: I've had trouble finding specific features in the ED for these puzzles, but the starting structures seemed to be in good places, so my work has been lots of recipe-refinement.
[10:20 AM] Susume: (she/her): I only tried 2100 and 2108, starting from an extended chain because I'm stubborn like that - 1st went great, 2nd I could not solve from ext chain
[10:20 AM] jeff101: It's been nice to get a string of top-10 solos.
[10:20 AM] jeff101: These are pretty rare for me.
[10:21 AM] jmbrownlee333: It takes away from the challenge in a good way. It is sort of like the final steps in refinement, skipping the laborius chain tracing.
[10:21 AM] beta_helix: I guess it's nice that you can always start from an extended chain if you want to (or are too stubborn )
[10:21 AM] beta_helix: @jeff101 I always expected ED puzzles to require a lot more hand-folding. I'm glad you are still able to use recipes with the density. Do you find yourself hand-folding more at the end? Or not more than usual?
[10:22 AM] jmbrownlee333: I would say not more than usual.
[10:23 AM] beta_helix: Interesting… Susume:, I assume you do a lot of hand work (hard to run only scripts from an extended chain!)
[10:23 AM] jeff101: almost no hand-folding on my part
[10:24 AM] beta_helix: and into the top-10 with no hand-folding… that's awesome!
[10:26 AM] beta_helix: This is a random question, but if we had only given you an extended chain, but the AlphaFold button was on, would you all use that as a starting model?
[10:27 AM] jmbrownlee333: YAh!!
[10:27 AM] jeff101: I think I would too.
[10:28 AM] Susume: (she/her): yeah, on ext chain I do hand work until it is all placed, then scripts, then another round of hand work later if I have time- if starting from a given model, I start by walking the whole chain looking for places where I think I can improve it by hand - which is not fun unless there are some obvious places. If I can't find visible flaws, it's all scripts from there - not my favorite way to play, but many others are great at it, so keep 'em coming!
[10:30 AM] Susume: (she/her): not sure I would use the foldit AF as a start, since it doesn't have coevolution data - that's fine for designed proteins but not great for natural ones
[10:30 AM] beta_helix: Thanks! When you say: "walking the whole chain looking for places where I think I can improve it by hand" do you mean manually, residue by residue… or looking at the individual score of the backbone/sidechains?
[10:30 AM] jeff101: if you will have externded chain as initial structure, please at least give us the Blueprint Tool to use.
[10:31 AM] jmbrownlee333: I walk pretty much residue by residue.
[10:31 AM] beta_helix: Don't worry, jeff101… we don't plan on not providing the structures. The goal of this is to see if Foldit players can improve on models that have been deposited in the PDB.
[10:31 AM] Susume: (she/her): manually, by eye. if I had time to play a lot of these though I would try some scripts that help to look at subscores
[10:31 AM] beta_helix: Got it… now I see how that is not fun!
[10:32 AM] jmbrownlee333: could we have a tool thats color-by-density-score?
[10:33 AM] jeff101: yes please … color by density score would be helpful
[10:33 AM] beta_helix: It would be pretty useful to be able to visually know which parts of the protein are not fitting well to density (we already can see those that don't fit at all)
[10:34 AM] beta_helix: I will bring your suggestion up at today's Foldit meeting
Thanks!
[10:34 AM] Susume: (she/her): are you going to check whether the areas that foldit players improve are the same ones that are flagged by the PDB's own quality metrics?
[10:36 AM] beta_helix: Yes indeed. In fact, there is already a website called PDBREDO that specifically focuses on improving the PDB. The cool thing is that, so far, the best Foldit models have been even better than PDBREDO!
[10:36 AM] beta_helix: We've already contacted the PDBREDO team and they seem interested in helping us out!
[10:38 AM] beta_helix: My student AJ (who you know, Susume) suggested: "Imagine if a scientist wanted to use a structure as a starting point for a new design problem, but they didn't really trust the PDB model on the RCSB website. How cool would it be if there was a "ask Foldit players for help" button, that would add it to the Foldit puzzle queue"
[10:39 AM] Susume: (she/her): I often look at the quality metrics on a PDB structure and think 'huh, ok, idk how to make use of that' - I'm glad they are there, but idk how to benefit from them. I LOVE AJ's idea!
[10:39 AM] beta_helix: It would kinda be the opposite of the "Share with Scientists" button
[10:39 AM] beta_helix: "Scientists asking YOU to share"
[10:39 AM] jeff101: the R and Rfree measures, what things within Foldit best reflect them?
[10:40 AM] Susume: (she/her): in Foldit land, scientists share with YOU!
[10:41 AM] beta_helix: @jeff101 Well, they represent the fit to density… so matching the density is important.
[10:41 AM] beta_helix: As I mentioned earlier, though, our long term plan (like… in 1-2 years) would be for Foldit players to recalculate the density based on your model and the phase information
[10:43 AM] beta_helix: essentially refining your density as often as you want, as you improve your model.
[10:43 AM] beta_helix: That's the idea, anyway
[10:43 AM] jeff101: would we be able to show our own latest version of the density while we work on a puzzle?
[10:43 AM] jeff101: like we calculate our own custom ED cloud?
[10:43 AM] beta_helix: You would be able to work with the density you more recently generated. Yes, exactly!
[10:44 AM] beta_helix: This is how crystallography programs do it.
[10:44 AM] beta_helix: They never come up with the perfect density and model on the first try.
[10:45 AM] beta_helix: Using the phase information, they generate density, refine a model into it, rephase, get a better map, etc.
[10:45 AM] jeff101: do you know of the ED clouds for these recent ED puzzles have been coarser than past ones? is it just me having trouble finding specific features?
[10:45 AM] petridecus: hi everyone, im andreas, the developer currently working on the trim tool for foldit! i can help answer any questions people have about it
[10:46 AM] Susume: (she/her): hi andreas! is this tool for trimming the density cloud?
[10:46 AM] beta_helix: This is why Foldit players have always impressed me with their ED work (when comparing to crystallographers). The expert in the paper I shared earlier was able to refine as much as they wanted… but Foldit players only had 1 map to work with!
[10:46 AM] petridecus: the trim tool is meant to reduce the active region of the protein that is being worked with @Susume: (she/her)
[10:47 AM] beta_helix: @jeff101 Yes, because these models are not great, the density hasn't been as high res as some of the previous ED puzzles
[10:47 AM] petridecus: you can use selection to isolate a certain region of a protein and then “trim” to work with that selection as if it were its own protein
[10:47 AM] beta_helix: https://fold.it/portal/node/2012556
Sneak Peak - The Trim Tool
Sneak Peak - The Trim Tool
[10:47 AM] beta_helix: (no density cloud there!)
[10:47 AM] petridecus: this will allow foldit to work with much larger proteins than before which could be especially helpful for ED puzzles
[10:48 AM] Susume: (she/her): oh ok cool!
[10:48 AM] jmbrownlee333: will we be able to control the slab depth more easily?
[10:50 AM] beta_helix: Andreas… how large a protein have you been able to run in Foldit without lag? (ie it almost feels like a 150-residue protein)
[10:50 AM] petridecus: with the trim tool you mean?
[10:51 AM] beta_helix: yes
[10:51 AM] petridecus: once you trim down to a region of N residues all puzzles effectively behave the same
[10:52 AM] petridecus: trimming and untrimming takes longer for super large proteins
[10:53 AM] beta_helix: @jmbrownlee333 can you describe "more easily"
[10:53 AM] Susume: (she/her): when you untrim, is there a cut at each end of your recent subset, where it connects back to the restored portion?
[10:53 AM] petridecus: yes, and in some cases there are a number of cuts
[10:54 AM] jmbrownlee333: I would kinda like a slider or dial to adjust slab width.
[10:54 AM] petridecus: also it’s worth noting that there will be an ensuing update where the “untrimmed” region i.e the residues beyond the selected region, will still be displayed (in a simplified fashion)
[10:55 AM] beta_helix: …so you can still see where the rest of the protein is, but it doesn't kill your rendering
[10:55 AM] jmbrownlee333: I like that idea.
[10:56 AM] beta_helix: Once the Trim Tool is in devprev, and any initial bugs resolved, Andreas will hold an Office Hour for your feedback, before releasing it to main. Maybe even having it open in Foldit over Discord to talk about specifics…
[10:57 AM] beta_helix: Andreas… I said "early next month" for its release. Is that realistic?
[10:57 AM] petridecus: definitely
[10:57 AM] jeff101: in the recent handful of ED Puzzles, have the best-scoring solutions agreed with each other?
[10:57 AM] beta_helix: (We won't release it on April 1st where you can only Trim, and unTrim doesn't work )
[10:58 AM] jmbrownlee333: lol
[10:58 AM] beta_helix: Any last questions for us today?
[10:58 AM] jmbrownlee333: When can we expect a blog post about how players are doing on these puzzles?
[10:59 AM] beta_helix: Great question! Probably after we talk to the PDB REDO and RCSB teams… by then we'll have a bigger sample size as well.
[10:59 AM] jeff101: also, see my question about 2 minutes ago
[10:59 AM] beta_helix: Great job already on the ED puzzles… very exciting! Can't wait to see what you can do with Andreas' new tools!
[10:59 AM] beta_helix: Sorry Jeff, missed that one…
[11:00 AM] Susume: (she/her): the RCSB team are among my heroes - collaborating with them would be so cool
[11:00 AM] beta_helix: @Jeff I believe the top solution has always been in the top5-10, but they have not all perfectly converged on the same topology
[11:00 AM] beta_helix: @Susume: (she/her) Right?!!
[11:01 AM] jeff101: thanks for having this office hour
[11:01 AM] beta_helix: Thank you all for stopping by today, especially on a weekday!
[11:02 AM] beta_helix: Thanks again and keep up the great folding!!!
[11:02 AM] beta_helix: Bye bye…
[11:02 AM] jeff101: bye

[agcohn821]

2:59 PM] jflat06: Hey everyone, I'll be around starting now for the office hours. I am a foldit developer with lots of experience with the code and a bit of biochemistry experience too.
[3:02 PM] Nicm25: hi! jflat06
[3:04 PM] jflat06: hello!
[3:15 PM] pc: hi
[3:15 PM] pc: what is the subject or kind of questions ?
[3:18 PM] pc: @jflat06 . Is it possible to have a buns detector tool in foldit, that is slower, but can detect without false positives (I mean not an objective but a separated tool) ?
[3:18 PM] jflat06: No specific subject, just as long as it's within my ballpark
[3:18 PM] jflat06: probably! I think we may even have such a thing
[3:19 PM] jflat06: It's always a big tradeoff between making things fast enough to be usable but still accurate enough to be useful
[3:19 PM] jflat06: We have a number of algorithms that people use internally in the lab for Rosetta that just aren't fast enough for foldit
[3:20 PM] pc: For me, AlphaFold is this kind of tool. a slow tool that make a protein Analysis. it is interesting to know if we are in a good or bad way
[3:21 PM] pc: as a developer to you watch "shared solution to scientists" ?
[3:22 PM] pc: @here office hours started
[3:23 PM] pc: what do you think about this suggestion : "A score modding for each objective". It can be usefull for buns for example. sometimes we need a different score penalty in each round phases.
[3:26 PM] pc: so "band recipes" or "mutate and wiggle recipes can be only to help "searching something" by trying multiple solutions. and "score modding" can be a setup that give player to goal to find. If we want less buns, if we want to increase surface..
[3:26 PM] jflat06: you mean additional options like "clashing importance"?
[3:27 PM] pc: yes, like the other under
[3:27 PM] pc: sidechain H-bond
[3:28 PM] pc: if I understand, a recipe think bonds give more points when it computes global score ?
[3:30 PM] pc: (with slider at higher value)
[3:31 PM] jeff101: I think the global score stays the same, but when you use tools like wiggle, they behave like the score being optimized has no clashing (for clashing importance of 0) or very strong hbonding (the hbond sliders set to 3)
[3:31 PM] jeff101: the global score shown to players stays the same
[3:31 PM] jflat06: the recipe uses the same score function as the one you see in the game, but it's possible for us to add additional sliders for various score terms. I think we have talked about this idea in meetings before, and it seemed like it wasn't quite as compelling as the case for clashing importance.
[3:32 PM] jflat06: but I think that's probably on a case-by-case basis for each score term, and it's possible that some of them might end up being very useful
[3:32 PM] jflat06: I can bring it up at our next meeting to get peoples' thoughts on it
[3:33 PM] alcor29: Hi jflat! 1. will the Graph Properties be restored so that my settings are saved? 2. Would it be possible to have a mutate program that can recognize that giving up a 600 point HBN for a 200 point segment is not a good idea, even for the integrity of the fold as well as for the global score?
[3:33 PM] pc: thanks
[3:34 PM] pc: yes, it can be usefull to have a easy control to what a recipe search, without modify script. (less buns, more surface, more ddg)
[3:35 PM] jflat06: 1. Can you clarify? 2. Mutate is supposed to do that, but it makes exceptions for if you have bands on your protein, or other situations where the score may technically need to go down.
[3:36 PM] pc: I mostly do hand folding, and recipes helps to find last optimisations. but it often not optimises what I want : with new buns score, now it add new buns each time I want to increase score with recipes.
[3:37 PM] jeff101: I have a question about puzzle 2137 (an ED puzzle). I am running it with clients from Oct 4 2021 to Jan 11 2022. I can save and share my solutions, but when I try to load my teammates' solutions, there is an error message and such solutions don't load. Did you guys change the internal file format for *ir_solution files recently so older clients cannot read ones made by newer clients?
[3:37 PM] jflat06: Internally, when you put bands on a model, the score function that the action is using is altered to include extra constraints. So it necessarily will lead to a decrease in foldit score, so we have to allow the score to drop in that case. However, this has collateral damage on things like your hbnet results.
[3:38 PM] jflat06: We have changed the solution file format recently, but it should be in a way that is fully compatible unless you have a very old client.
[3:38 PM] alcor29: Would this last change explain why it's harder to make and hold HBNs?
[3:39 PM] jflat06: I dont think so, though there some changes a few releases ago that modified the way these work, if i recall correctly
[3:41 PM] alcor29: @jflat06 Clarification of 1. On the old UIs i would set the number of undos that I wanted and the RAM and these settings were saved. Since the new UI these are not saved and I have to modify for every client I start.
[3:41 PM] jflat06: oooh, i see
[3:42 PM] jeff101: I think ichwill requested that hydrogen bond networks be calculated differently so more AA combinations could give 100% bonding
[3:42 PM] jflat06: that sounds like an issue for @Sciren (he/him) - I'd recommend making a feedback about it. We modified the ways settings are saved, and there's been some work trying to get everything updated to the new system.
[3:43 PM] jflat06: If you see anything like that, we can fix it, just let us know!
[3:45 PM] jeff101: @jflat06 what features have you been working on lately? what can you tell us about them?
[3:47 PM] alcor29: Maybe it's just me but lately it seems HBNs harder to achieve and hold. Will make a feedback for @Sciren (he/him) re graph properties.
[3:48 PM] jflat06: I've been helping with some exciting new features from bkoep, but I don't want to steal his thunder about it, so I'll just say that you guys are going to love it
[3:49 PM] pc: I suggested a new objective for binder puzzle that add penalty with solutions with too mush void in contact surface. But maybe it is better to have a better CS objective that take count about thoses voids.
[3:50 PM] jflat06: doing anything with voids is tricky, because our algorithm for detecting them isn't very good
[3:50 PM] jeff101: is covid still keeping most of the lab working remotely? are more people doing experiments in the lab these days?
[3:50 PM] jflat06: I think most people are actually back in the Baker Lab and doing experiments
[3:51 PM] jflat06: I actually work in the Comp Sci department, so I don't go into that lab other than for the occasional meeting
[3:51 PM] pc: "void resolution" looks to be important to have a successful binder protein.
[3:51 PM] jeff101: are the meetings more in-person these days?
[3:52 PM] jflat06: some of them have been, though we haven't been doing it as much recently.
[3:52 PM] jflat06: I don't think there's much in the way of covid restrictions affecting the lab anymore
[3:53 PM] jeff101: that's good
[3:53 PM] jflat06: re: voids - there's actually a much more accurate algorithm called dAlphaball
[3:53 PM] jflat06: but it's written in FORTRAN, if i recall correctly
[3:53 PM] pc: oh nice
[3:53 PM] pc: it can be good to have a more precise visualisation of them too ^^
[3:53 PM] jflat06: so it is difficult to integrate into Foldit, and it's slow…
[3:54 PM] alcor29: Having flashback to the 50s……….Fortran still around..
[3:54 PM] jeff101: how about the new Foldit web page? are you working on that? any estimate when it will go online? weeks? months? years?
[3:55 PM] pc: I try to pack my core inside my protein perfectly, or have a better contact interface, but i can't visualise all (I have only some red spheres sometimes).
[3:55 PM] jeff101: there are newer versions of fortran called fortran 77 and fortran 90
[3:55 PM] pc: thanks, can be interesting to have better void detection and visualisation ^^
[3:55 PM] alcor29: TX Jeff.
[3:56 PM] pc: of course a slow tool that is not an objective can be interesting too !
[3:56 PM] jflat06: The website should be coming 'soon' - it gets put on the backburner every time something more pressing comes up, like the work on Alphafold or some of these new tools that we have in the mix.
[3:57 PM] alcor29: I hope the new recipe listings will be easier to search, and will respond to words in the middle of the description.
[3:58 PM] jflat06: I believe we have that implemented! At the very least, making changes on the new site code is much easier than for the existing site, so feature requests like that are much more viable to implement.
[3:58 PM] Nicm25: For voids: it's enough to visualize them as densities in metapolygons just NOT the electron density as usual.
[3:59 PM] jeff101: @alcor29 if a recipe is already in your cookbook, you can look in the cookbooks all.macro file to search for special bits of code
[3:59 PM] alcor29: I mean the web listing not the client list.
[4:00 PM] jflat06: Yeah, the new web interface for recipes is more fully featured than the existing site, so hopefully you'll be pleased.
[4:01 PM] alcor29: @jeff101 Thanks. That's interesting. Though I once got into trouble by opening the all.macro file. Was advised never to open it.
[4:01 PM] jflat06: Well the all.macro file will be a thing of the past. Instead you'll have an extra 'recipes' directory that has a separate file for each recipe, all written in JSON
[4:01 PM] jflat06: should be a lot easier to work with
[4:02 PM] jeff101: I sometimes look at the all.macro file using Notepad or Wordpad. It is safe as long as I don't change it and then save it.
[4:02 PM] jeff101: When I am coding, I sometimes look for examples of how to use certain LUA commands by searching the all.macro file.
[4:02 PM] alcor29: Got it. Thanks jeff101
[4:04 PM] alcor29: Thanks @jflat06 . Appreciate it.
[4:05 PM] Nicm25: Thanks, new question to ask this, when editing recipe file do We need to calculate the verify? or do We need to define it in game as my own user?
[4:05 PM] jflat06: In the new system, you'll edit the files on the website, and then just click 'refresh' to load your changes into the game.
[4:06 PM] jflat06: There's also a button that lets you load a raw text file as a recipe, too
[4:06 PM] Nicm25: thanks for answering and thank you for today.
[4:07 PM] jflat06: No problem! I am going to go find some lunch now. Thanks for coming, everyone!
[4:07 PM] pc: thanks ^^
[4:07 PM] jeff101: thanks to you too jflat06

[agcohn821]

10:02 AM] petridecus: hey everyone! i’m hosting an office hour right now, the topic of which is feedback and overall thoughts on the new trim tool
[10:02 AM] petridecus: looking forward to hearing about everyone’s experiences with it thus far and how we can improve it
[10:07 AM] alcor29: Good morning petridecus. Say I choose to divide the protein into 4 sections, and I work a little on each. Then I want to go back to section 3 to do something else. I don't undserstand how to toggle or select a particular section. It seems to choose whichever section it wants to choose.
[10:08 AM] petridecus: hi @alcor29 thanks for the question! at the moment the best way to do that would be to just untrim, select a new region of residues, and trim again. do you think it would be beneficial to be able to jump directly between trimmed regions?
[10:09 AM] alcor29: I think so. It is frustrating to have to keep reselecting a section each time.
[10:10 AM] petridecus: would you like to be able to save/keep track of what the selection was for different trim regions you’ve worked with?
[10:11 AM] alcor29: Yes.
[10:12 AM] petridecus: ok. i think that’s a great idea and will bring it up with the team at our next meeting
[10:12 AM] petridecus: out of curiosity has anyone played with the new trim puzzle in dev prev yet and have any feedback on the new visualization which was just added?
[10:13 AM] cjddig: (he/him): Hello petridecus, I'm using devprev.
[10:14 AM] cjddig: (he/him): However I do not use main version, so I do not know the difference between devprev and main.
[10:15 AM] petridecus: i could be wrong and the new feature hasn’t been added yet but, there should either already be or soon be a new minimal visualization of the full protein even when working within the trim tool
[10:16 AM] cjddig: (he/him): Do you mind if I send the screenshot of mine?
[10:16 AM] petridecus: sure!
[10:16 AM]
BOT
FolditBot: «b>cjddig:</b>> IMAGE: http://fold.it/portal/files/chatimg/irc_869190_1654445782.png
Image
[10:16 AM] petridecus: ah ok, it’s not there yet!
[10:17 AM] petridecus: so i guess that new visualization is just something upcoming for dev prev players to look forward to
[10:17 AM] cjddig: (he/him): I wonder how it'll look. Anyway, I have another question. when I use trim tool and untrim, then too many clashes appear. Is trim tool optimized for hand-folding? Should I do hand-folding, not using wiggle or shake?
[10:18 AM] petridecus: so this is something we’ve started to notice in feedback with the tool thus far and anticipated was possible (and which should be quite easy to fix)
[10:19 AM] petridecus: it seems like when wiggling in the trim pose, issues with the backbone are created at the edges of the trim region
[10:19 AM] petridecus: one workaround could be to just select the full trim region EXCEPT for the edges and just work on that
[10:19 AM] petridecus: that will avoid creating any weird backbone issues once you untrim
[10:20 AM] petridecus: there are currently measures in places to hold those edges where they are but there seem to be exceptions where they don’t work perfectly
[10:20 AM] alcor29: Will there ever be a use or need for scripts in these puzzles?
[10:21 AM] alcor29: Or just wiggle and shake when done with the hand folding.
[10:21 AM] petridecus: hmm i’m no expert on scripts in foldit but it should be no different from how scripts are used in other puzzles to my knowledge
[10:21 AM] cjddig: (he/him): I usually use sphere selecting and trim, so it's sometimes uncomfortable to select the segments except edge ones.
[10:22 AM] petridecus: yeah i agree it’s a hit to efficiency and not fool proof so it’s far from ideal
[10:22 AM] petridecus: but it could help in the short term as an option until we roll out an official fix!
[10:23 AM] petridecus: or also just freezing those edge residues might help (not something i’ve tried yet though)
[10:25 AM] cjddig: (he/him): 'Idealize peptide bonds' tool paralyzes all endeavors not to make backbone issues… I think we should not use too powerful tools like idealize tool.
[10:25 AM] petridecus: within a trimmed solution you mean?
[10:25 AM] cjddig: (he/him): Yes
[10:26 AM] petridecus: yeah i think you just need to use discretion with certain tools within the trim tool
[10:28 AM] petridecus: with a better system for locking edges and this new visualization in place though that should become much easier
[10:28 AM] cjddig: (he/him): I just found cutting the edge segments might help. I want that new feature you mentioned so that I can predict whether the modifications I made would cause clashes.
[10:28 AM] petridecus: yeah i think it’ll be there whenever the next dev prev release comes out
[10:29 AM] Nicm25: hello!, He was curious about scripts, which may duplicate what folder has already asked so far, but what I think is a bit long winded…
[10:29 AM] petridecus: hi @Nicm25 i’m not sure i’d be the most helpful in answering a script question but i could try my best
[10:31 AM] Nicm25: From my point of view as writer of recipe will simply be question, write when if other users not have question.
[10:32 AM] Nicm25: I thought that even from Recipe writer's point of view, it was necessary to check if is trimmed or not.
as pointed out earlier, when I was trying to create a tool that would automatically change location of residues that should be trimmed, I sometimes wanted to know extent to which they had already been trimmed.
I believe these improvements would be difficult to make. Does Rosetta functions only take into account residues that are currently displayed? If so, does it get any information at all about hidden residues excluded by trim tools?
[10:33 AM] petridecus: yeah that’s pretty much the purpose of the tool. only the trimmed residues are operated on
[10:34 AM] cjddig: (he/him): Oh, then the recipe can't select the residues which are not trimmed?
[10:34 AM] petridecus: yeah i highly doubt it
[10:34 AM] petridecus: i know next to nothing about scripts but i don’t see how that would be possible in a trimmed puzzle
[10:35 AM] petridecus: when you trim, foldit effectively sees an entirely new protein that contains only those residues until you untrim
[10:35 AM] alcor29: Are you working to fix the crashiness? I have crashed 3 times since we have been talking. Changing tracks and moving the cursor once too many times.
[10:36 AM] petridecus: that’ll be at the top of my agenda for the week. im struggling to understand whether this bug is specific to the trim tool or if it’s also OS specific
[10:36 AM] alcor29: Also where is the best place to report issues to you?
[10:36 AM] petridecus: so it’s a bit hairy. i haven’t been able to recreate it on my linux machine!
[10:36 AM] petridecus: the bugs and feedback channel in discord!
[10:36 AM] alcor29: k
[10:37 AM] Nicm25: what we can see from recipe in that if all we have is residues currently displayed, then we have no way of knowing where trim started or where it ended.
because if we have no way of knowing if we even got residue number, it means we won't be able to achieve automatic control…
[10:37 AM] Nicm25: Anyway, thanks for answering.
[10:37 AM] petridecus: so you want to know where all the current residues exist relative to the full protein?
[10:38 AM] petridecus: under the hood we keep track of a “mapping” of sorts that relates the trim region to the full protein. perhaps it could be helpful for these script issues as well
[10:38 AM] petridecus: i can ask the team about this
[10:39 AM] Nicm25: ex… if can achieve that, it can detect trim edges and prevent unnecessary optimization, crashes. :)
[10:39 AM] petridecus: got it, yeah i can see how that would be helpful
[10:40 AM] cjddig: (he/him): Another question, if I trim 131st residue, is its residue number 131 or 1? Currently turned off my computer, so I can't check this.
[10:42 AM] alcor29: I ran a recipe yesterday and it read as 1
[10:42 AM] petridecus: yup that sounds right
[10:42 AM] alcor29: Sorry, not wide awake yet. Can't remember wich script.
[10:42 AM] petridecus: cause again, it’s effectively seen as its own unique protein under the hood while it’s active
[10:43 AM] petridecus: that bit of insight might help in understanding certain aspects of the behavior
[10:43 AM] petridecus: in this case that means it’s counting residues from 1 as opposed to the number from the original protein
[10:44 AM] petridecus: hope that makes sense
[10:44 AM] cjddig: (he/him): I got it
[10:46 AM] cjddig: (he/him): Then, does foldit see a multiple pieces of protein if I selected multiple separate segments?
[10:47 AM] petridecus: nope it’s all just one structure with space between
[10:47 AM] petridecus: (not multiple proteins)
[10:47 AM] cjddig: (he/him): Oh I see
[10:47 AM] alcor29: (I just crashed again twice once because I was merely trying to translate the protein in space. So currently it may be unworkable for many folders. )
[10:49 AM] cjddig: (he/him): Actually 2155b per se seems to cause a lot of crashes
[10:50 AM] alcor29: BTW. I almost never crash.
[10:51 AM] petridecus: what operating systems are you guys using?
[10:51 AM] alcor29: Win11 64bit
[10:51 AM] cjddig: (he/him): I use Windows 11.
[10:51 AM] petridecus: yeah i think the bug is specific to windows. quite annoying for me as a developer who pretty much only ever uses linux haha
[10:52 AM] Nicm25: Linux Ubuntu and CentOS amd64 :)
[10:53 AM] cjddig: (he/him): Maybe I'll play foldit on VM soon, to make more helpful bug report to you
[10:54 AM] Nicm25: my problem is that I don't have a lot of available resource allocations, so I don't have opportunity to report bugs.
[10:55 AM] cjddig: (he/him): Anyway, my last question: do you think trim tool will change the ED puzzle, or ultimately, Foldit and the way to fold?
[10:57 AM] alcor29: Thanks petridecus!

[agcohn821]

1:00 PM] rmoretti: Hello! I'm rmoretti, a Foldit developer, and I'm here for Office Hours. I'm happy to answer any questions you have about small molecule design, the Rosetta engine behind Foldit, or any other questions about Foldit you may have.
[1:16 PM] sandrix72: Hi, Moretti
[1:16 PM] alcor29: Hi Moretti
[1:16 PM] rmoretti: Hi!
[1:16 PM] alcor29: Not many here today. Maybe they mixed up the time like I did.
[1:16 PM] sandrix72: May i have some questions regarding several different duties regarding foldit?
[1:17 PM] rmoretti: It's also a holiday weekend in the US.
[1:17 PM] alcor29: I have no questions. Maybe you have some
[1:17 PM] rmoretti: Feel free to ask any questions about Foldit. I might not have all the answers, but I'll try.
[1:17 PM] sandrix72: ok
[1:17 PM] rmoretti: I wouldn't mind hearing your thoughts about the small molecule design puzzles, if you have any.
[1:17 PM] sandrix72: 1) Will be in the close future ligand modelling puzzles?
[1:19 PM] rmoretti: The hope is to continue to run small molecule design puzzles. We have a few potential targets we're thinking about at the moment, though it may be a couple of months until we post the next ones.
[1:19 PM] sandrix72: 2) If yes, there is a will to create functions that may add/delete atoms via a recipe?
[1:19 PM] Susume: (she/her): here's a question from spvincent, who can't make it to ofc hr: "I'm wondering how many submitted solutions contain atoms other than CON. The interface allows you to add things like CF3 groups but despite the fact that these seem to occur regularly in drug molecules there always seems a penalty when you try to use them in Foldit."
[1:19 PM] sandrix72: i expect using such functions we could create better designs
[1:21 PM] rmoretti: We've been thinking about adding small molecule design-specific lua functions, but adding those is a bit lower on the list, as we're hoping the tap human creativity for design, rather than turning it into a game of automated dart throwing. (There's already some algorithms which take the automated random approach, so we're hoping to leverage what makes Foldit unique.)
[1:22 PM] sandrix72: I want to say, that simply adding atoms we will not obtain a good result, creativity is still needed
[1:23 PM] sandrix72: for example the "where" to put is not creating any distorsion is crucial
[1:23 PM] rmoretti: Regarding atom types, there's no specific penalty for non CON atoms (aside from the atom count objective on the recent KLHDC2 puzzle series.) That said, the other atom types are only going to make sense if they fit energetically with the rest of the protein. I haven't looked at the rate of non CON atoms in the results – I think there are some, but possibly not many.
[1:23 PM] sandrix72: which cannot be solved "simply" by an automata
[1:23 PM] alcor29: Also medicines seem to have the acid to help break them down built in. I see many X hydrochlorides. When I try to add Cl though scores always go down.
[1:24 PM] rmoretti: But it's also true that adding other atoms is sometimes a late stage optimization in "normal" drug design – early stage compounds do tend to be CON heavy, even in non-Foldit contexts.
[1:25 PM] sandrix72: there will be in the close future "Godzilla" proteins?
[1:25 PM] alcor29: *CL
[1:26 PM] rmoretti: The "drugname hydrochloride" normally indicates that the chlorine is not part of the drug itself – it's just that the chloride is a counterion which is neutralizing a charge on the active molecule. When you put it in water, the chloride dissociates and goes off on its own. In Foldit we just model the active species, so you're not going to see those sorts of counterion chlorines. (But you will see the positively charged nitrogens which they are countering the charge of.)
[1:26 PM] sandrix72: Windows based PC-s cannot handle larger proteins than about 200 segments without rewriting the recipes eliminating for example all wiggleall instructions
[1:27 PM] alcor29: Thanks.
[1:28 PM] rmoretti: The new trim tool has been developed to help deal with some of the larger proteins, and we're hoping it will enable having more "Godzilla" proteins, especially in the electron density context (where modeling large proteins is important). We're still working out details on it, though, and being able to make it usable on most machines.
[1:29 PM] sandrix72: Trim has serious problems
[1:29 PM] rmoretti: But, yeah, you're likely going to need to come up with a workflow which is "trim aware" – I think you can run "whole protein" recipes when you have a trimmed structure and the recipie will only act on the trimmed portion.
[1:30 PM] sandrix72: in case of huge proteins i would like to suggest some observations (i am not sure that implementing thaem is possible or not)
[1:30 PM] sandrix72: regarding trim
[1:30 PM] alcor29: So far Trim is no help to me. Can work the whole protein though as before.
[1:30 PM] sandrix72: 1) we should select a "Cube" instead a segment
[1:30 PM] alcor29: Which may be what is happening.
[1:31 PM] sandrix72: 2) Automatically freeze all segments in the outer surface of the cube
[1:32 PM] sandrix72: 3) If it is possible, permit merging solutions using many cubes at the same time on different thread
[1:32 PM] rmoretti: There's already a "sphere select" mouse shortcut - it should be listed in the help, but I think it's something like ctrl-shift-drag. That is one of the major ways the trim tool is anticipated to be used. (Though I agree that functionality is perhaps not as well documented as it should be.)
[1:32 PM] alcor29: The main prob seems to be that whatever progress was made in the trimmed part, it all changes when the whole protein is manipulated.
[1:33 PM] sandrix72: As long as the external part is frozen, the merging theoretically is possible
[1:33 PM] rmoretti: I know that there's some tweaks in the work to make the interface between the trimmed and the non-trimmed portion work better, such that you don't mess things up when you untrim. (Going toward the better "freezing" behavior.)
[1:34 PM] alcor29: But, as mentioned before it's going to take a whole while to get used to it.
[1:35 PM] rmoretti: Yeah, its a tool still in the early phases, and there's still heavy development on making it work better. If you have suggestions or complaints, please be sure to open a Feedback item on them. (We do look at all of the Feedbacks, even if we don't necessarily act on them immediately.)
[1:36 PM] sandrix72: Breafly i described by idea, using cubes instead of segments, permitting in this way developing multiple cubed parts if they have no intersection
[1:37 PM] sandrix72: and to have foldit command handling the trim via recipe
[1:38 PM] rmoretti: I think the goal is to have it such that you can trim a section, manipulate it then untrim it such that it "seamlessly" merges back to the full structure, then you can make a new selection, trim to it, work on it in isolation and then merge it back. So you should be able to merge different sections.
[1:38 PM] alcor29: Yup.
[1:39 PM] rmoretti: I think being able to trim via recipes is on the books , but I think the priority is getting the interactive version to work better first. (But feel free to open a feedback if you think it should be a higher priority)
[1:41 PM] sandrix72: ok, but a segment would be much harder to insert back into the whole protein seemlessly than a "cube"
[1:41 PM] sandrix72: functions have no priority until trim works well
[1:42 PM] rmoretti: Yeah, there's some flexibility lost if you only select a single segment and trim that – but you should be able to select larger sets of segments (secondary structure elements and a sphere of surrounding segments) and those should all be included in the trim.
[1:43 PM] alcor29: Do you think that eventually folders will be able to use P, S or Cl, Fl etc. while still playing the score game? Or will we just stick to CON due to the scoring necessity? Will there be any changes to the scoring?
[1:43 PM] sandrix72: what is the maximum segment number which we should handle in mid term?
[1:45 PM] sandrix72: At ligand i used almost only C, scoring permitted nothing else
[1:45 PM] rmoretti: I think that P,S, Cl and F should be usable in the right circumstances. You are correct that the current scoring hasn't really been optimized for those sorts of non-protein atoms (the most work on the score function is for protein interactions). We may need to come back and tweak how those atom interactions are scored for better results.
[1:47 PM] rmoretti: You'll definitely want to use more than just C, as you'll need O & N to make hydrogen bonds. Also, if you just use C you'll keep things too hydrophobic and your designs will either fail the objectives or will be tossed out.
[1:47 PM] alcor29: I' assuming that you expected that because you included all those atoms in the widget.
[1:48 PM] rmoretti: Compounds which are too greasy (e.g. all C) is something we've noticed in some puzzles, and which we've tried tweaking the objectives to address.
[1:49 PM] alcor29: So most of the near work on it will probably in tweaking the filters.
[1:49 PM] rmoretti: We certainly want players to have the potential to use those atoms – they are seen in commercial drugs, and they're often an important factor in the activity – but they may have to be used judiciously, and they're not always going to score well.
[1:49 PM] sandrix72: We were "forced" by the scoring system to use them
[1:50 PM] rmoretti: Yeah, a large amount of the effort in the small molecule design puzzles has been attempting to get a set of objectives which match the non-binding energy criteria which our experimental collaborators are using to determine whether or not the compounds are worth synthesizing.
[1:52 PM] rmoretti: (There's a large number of considerations which go into drug design, over and above the basic "does this molecule bind well to its protein target".)
[1:53 PM] sandrix72: From scientific point of wiew how large proteins we should handle?
[1:54 PM] rmoretti: There are some monster protein complexes out there – look up the ribosome and the nuclear pore complex to see how big they get. Ideally, we'd like Foldit to be able to handle all of those.
[1:54 PM] sandrix72: Sorry, i am computer scientist
[1:55 PM] alcor29: We'll need them to spring for some Crays.
[1:55 PM] rmoretti: In practice, how large of systems we run will be based on how good players' computers are at handling things, and how easy we can make the trim/untrim workflow.
[1:55 PM] sandrix72: Whast is the max number of segments?
[1:55 PM] sandrix72: considering an ideal case
[1:56 PM] rmoretti: I think the nuclear pore complex is about 125 MDa, which works out to over a million segments.
[1:57 PM] sandrix72: So we should handle 1M segments from scientific point of wiew
[1:57 PM] rmoretti: We're probably not going to give players something that big, but we certainly want to be able to work with things in the thousand-plus segments area. (Versus the current ~200 segment limit we're dealing with now.)
[1:58 PM] sandrix72: if you want so, i 0may have some suggestions how to solve them even using nowaday's PCs
[1:59 PM] sandrix72: Just from curiosity
[1:59 PM] sandrix72: that 588 segment solution was good enough?
[1:59 PM] alcor29: is this amount for active residues only or are the symmetry segments included.
[1:59 PM] rmoretti: If you have ideas, feel free to write up a Feedback. They may or may not work out in practice, but we will read it and think about it.
[2:00 PM] sandrix72: where to write it, in the forum?
[2:00 PM] rmoretti: I think the size of the nuclear pore complex includes all the symmetric partners.
[2:00 PM] sandrix72: i am new here
[2:00 PM] rmoretti: There's a Feedback tab on the current website: https://fold.it/portal/feedback
[2:01 PM] alcor29: So when we do pentamers for example we can handle 400 segs without much slowdown.
[2:01 PM] sandrix72: ok, thank you
[2:02 PM] sandrix72: those task are VERY slow on my computer, but symmetry puzzles are still a mistery for me
[2:02 PM] rmoretti: The way symmetry is handled is sort of like how the trim tool works. (Working with the main sense but getting the details completely wrong.) The "active" part of the system is only the asymmetric unit, and there's some shortcuts we can take internally which can basically copy over info to the symmetric partners, speeding things up.
[2:03 PM] sandrix72: So the 588 segment solution was good enough ?
[2:03 PM] alcor29: Hmmmmm
[2:04 PM] rmoretti: So even though there are ~400 segments in total, the performance is more along the lines of the ~100 segments that are in the asymetric unit. (Not exactly, but roughly.)
[2:04 PM] alcor29: Got it.
[2:05 PM] rmoretti: How many segments you need depends a bit on the system in question. Some proteins are small. Others are bigger. For some systems we can pre-trim things and only give a puzzle with part of the system. Others really need to be considered as a full system.
[2:06 PM] rmoretti: Any final comments on the small molecule design puzzles or general questions before I log off?
[2:06 PM] alcor29: Thanks moretti. Very instructive. Sorry about the turnour.
[2:07 PM] sandrix72: Thasmk you, i just wanted to know if we yieled good enough solution for the 588 segment protein,or not?
[2:08 PM] rmoretti: Sorry, I'm not involved with analyzing the results of that puzzle, and I can't recall hearing anything about it in a meeting.
[2:09 PM] alcor29: You all have a nice Holiday.
[2:09 PM] sandrix72: ok, thank you
[2:09 PM] rmoretti: Thanks for coming to the Office Hours, and have a good weekend!