Office Hours

[agcohn821]

[12:00 PM] horowsah: Hi all- I'm here for office hours for the next hour if you have anything you'd like to chat about. I should say that my expertise is on electron density puzzles, education, and general biochemistry. So design is not my strong suit, for example.
[12:11 PM] Nicm25: Hmm, IRC online? (ping)
[12:12 PM] jmbrownlee333: I have something i don't fully understand on ED puzzles. sometimes residue has clear density for one position, but the highest scoring foldit models show it doing something else. Case-in-point R68 in 2170(lysozume)
[12:12 PM] jmbrownlee333: lysozyme
[12:13 PM] horowsah: Yep, its a confounding thing in electron density that that can happen
[12:13 PM] horowsah: in the early days of crystallography, people would generally always follow the density
[12:14 PM] horowsah: but by doing so they would often break the laws of chemistry and physics and actually be fitting an experimental error
[12:14 PM] horowsah: so now the algorithms that are used try to weigh the experimental data against whether the data at that point could be a mistake
[12:14 PM] jmbrownlee333: So I assume we believe the ED more than foldit? Or are our phases not that good?
[12:15 PM] horowsah: Our phases vary pretty considerably, but only the very very very best data can make us reconsider what physics suggests
[12:15 PM] horowsah: and we almost never fall into that category
[12:16 PM] horowsah: so it's possible that when the density leads us to make something that is less energetically favorable it's correct, but we usually need extremely good data to make that the correct choice
[12:16 PM] horowsah: Of note, crystallography software has gotten pretty good at this, but cryo-EM software is not as good yet
[12:16 PM] jmbrownlee333: But lysozyme should be pretty awesome data, right?
[12:17 PM] horowsah: it depends- there are some awesome lysozyme structure out there, and some terrible lysozyme structures out there. Each crystal is different, and sometimes the same protein in different crystals acts a bit differently.
[12:20 PM] jmbrownlee333: So when playing foldit, its hard to know when to just let the algorithm rub free, and when to be a slave to the ED.
[12:20 PM] jmbrownlee333: run free
[12:20 PM] horowsah: Yes, it is hard. Generally, the guidance is two-fold. First, if the map and density is high quality, it's more likely to be good to let it match the density.
[12:20 PM] horowsah: The worse the map quality, the more you should let it run free.
[12:21 PM] horowsah: The second guideline is that to help scientists the most, it's best to try h. We will be able to see the differences and it might be that the thing that gets scored lower in Foldit might actually turn out to be slightly better after doing more follow-up calculations.
[12:22 PM] horowsah: On the first guideline, the map quality will vary at different positions in the map, so some spots you'll want it to match the density quite closely, and some you'll probably not want to listen to what the map says
[12:23 PM] jmbrownlee333: Ya, I think somebody said we might get to use our foldit solutions to update the phases/maps in the future. If I am remembering correctly.
[12:24 PM] horowsah: That's a goal we have- still needs a lot of programming work, though.
[12:27 PM] jeff101: Hi you all. I put several questions on the web page below:
[12:27 PM] jeff101: https://fold.it/portal/node/2013381
[12:27 PM] pc: https://fold.it/portal/files/chatimg/irc_959223_1657558127.png Hi @horowsah horowsah. In design puzzles, we mostly have a large hydrophobic zone, in contact zone, in our top score solutions. Is it a problem for lab tests ? Should players care about this, and use less hydrophobics ? Or a new objective can be needed for this ?
Image
[12:28 PM] horowsah: @pc I'm not really sure how bit is an issue. I'm going to tag @bkoep here for that one as I'm not sure whether that size is an issue
[12:30 PM] horowsah: Jeff, I'm not sure on the cis bond question.
[12:30 PM] pc: Thanks ^^ @horowsah. A second question : about AlphaFold : in prediction we often see that helixes are very straight. If they are a little twisted in Foldit, they become straight in AlphaFold each times. Is it an AlphaFold imprecision , or we really need to do very straight helixes in our designs ?
[12:30 PM] horowsah: My guess is that Foldit uses a blanket way of preventing cis bonds that's probably far from perfect
[12:31 PM] jeff101: Foldit puts yellow balloons near cis bonds to warn us about them. If you click the balloons, the cis bonds get fixed.
[12:31 PM] jeff101: Also, some tools will change cis bonds to trans ones automatically.
[12:32 PM] horowsah: As to whether a cis bond is more likely before or after a proline, that's also something I'm not positive on. I'm guessing someone out there knows the answer to it, but I don't off the top of my head. It probably depends strongly on context as well.
[12:32 PM] jeff101: It would be nice if we could change trans bonds to cis ones with the click of a button.
[12:33 PM] horowsah: Probably an alphafold imprecision. There's lots of little things Alphafold doesn't get right but are relatively easy to fix. The goal of Alphafold is to get global things right, figuring that it's easier to fix smaller local issues.
[12:33 PM] jeff101: Do you think most electron density fitting software ignores cis peptide bonds and forces them to be trans?
[12:34 PM] horowsah: So the most popular software used for crystallography by scientists is called Coot, and I think it lets you control cis to trans and back pretty easily, much more easily than Foldit does.
[12:34 PM] jmbrownlee333: preceding prolines generally, I believe
[12:34 PM] pc: thanks
[12:34 PM] horowsah: This is something we could probably improve
[12:35 PM] jeff101: so you think thr307-pro308 is more likely to be cis than pro307-thr308 ?
[12:35 PM] jmbrownlee333: yes
[12:36 PM] jeff101: is the "top residue" always the higher #'d one?
[12:36 PM] horowsah: I'm not really familiar with the top residue tom residue way of talking about it, actually
[12:36 PM] jeff101: which terminal is the "top residue" closer to? the NH2 terminal or the COO one?
[12:37 PM] jeff101: jflat06 said "top residue" in a post from 2013:
[12:37 PM] jeff101: https://fold.it/portal/node/995580#comment-24061
New Visualization Options
New Visualization Options
[12:38 PM] horowsah: Hmm, I think I'd need a picture with top and tom labeled to figure out what he was referring to
[12:39 PM] pc: @horowsah In protein reviews, bkoep often talks about confidence and never about similarity in AF. Maybe similarity is not enough precise, and we should not over optimise it ? (75% similarity with high confidence (85%) maybe is enough ?).
[12:39 PM] horowsah: I'm not fully confident on the numbers, but that seems to jive with what I remember.
[12:42 PM] spvincent: I was wondering about that in symmetry design. What use is high confidence without high similarity? The Hnets wouldm't have much chance of forming.
[12:42 PM] horowsah: @spvincent I'm going to tag @bkoep on that one as well, as I'm really not sure how that all plays out
[12:43 PM] pc: yea HBN are very precise to score well, so if similarity is too low.. maybe it fails
[12:43 PM] spvincent: tx
[12:44 PM] jeff101: are there plans to have the Trim Tool preserve bands and frozen residues?
[12:44 PM] spvincent: Any news on the small molecule design puzzles? Will there be more rounds? Did they find some likely-looking candidates and are busy synthesizing them?
[12:44 PM] horowsah: I haven't heard back yet on the bands and frozen residues, but I hope that's doable. I'll put a reminder to ask that next time I see the person working on improvements to the Trim tool.
[12:45 PM] jeff101: https://fold.it/portal/node/2013141#comment-46429 asks similar questions
Introducing the Trim Tool!
Introducing the Trim Tool!
[12:45 PM] jeff101: https://fold.it/portal/node/2013141#comment-46429
Introducing the Trim Tool!
Introducing the Trim Tool!
[12:45 PM] jeff101: are there plans to make LUA commands for Trim and Untrim?
[12:45 PM] horowsah: On the small molecule design puzzles: there's a lot of legal mumbo jumbo surrounding those and I think I might not be allowed to say anything yet, sadly. Hopefully soon.
[12:46 PM] horowsah: Lua for trim and untrim are I believe currently being worked on
[12:46 PM] jeff101: That's good. LUA commands for trim and untrim would be helpful.
[12:47 PM] horowsah: Yeah, scripting trim and untrim could make it feasible to do a lot of things not currently daoble
[12:48 PM] jeff101: I agree.
[12:50 PM] jeff101: I am curious how Foldit decides which peptide bonds to mark with yellow or red balloons. What does it do if a peptide bonds would be h yellow and red? Does it show as yellow or as red?
[12:51 PM] horowsah: I have no idea how it would choose in those cases. I hope it would choose wisely somehow, but not sure. I'm going to tag @jflat06 on that one because he might know the answer to that one.
[12:51 PM] jeff101: I've looked at some distances for bonds Foldit marks as yellow (cis) and found strange results. It makes me wonder if Foldit is doing it right.
[12:52 PM] horowsah: It could be something is up with it grouping other things in that category and calling it yellow.
[12:52 PM] jeff101: I've been focusing on puzzles 2158 and 2167 for these cis bond measurements.
[12:53 PM] jflat06: I'm pretty sure yellow denotes cis bonds and red is ideality issues
[12:53 PM] jeff101: If I shared some solutions with scientists and e-mailed you about them, would you be able to check them?
[12:53 PM] jeff101: what is a bond is h cis and has ideality issues? what color will it get?
[12:53 PM] horowsah: I think so- I don't get those by default but if you sent me a message about them when you do someone can track them down and route the mto me
[12:54 PM] jflat06: If h are present, my guess is h are being rendered and it is undefined which you'll see
[12:54 PM] horowsah: Personally, if it was cis and had ideality issues I'd prefer it show red, but I don't know which it actually is
[12:54 PM] jflat06: Yes, it should show red
[12:55 PM] jeff101: but if it is really cis, I think that wiggle will keep it cis
[12:55 PM] jflat06: Red should always be fixed, whereas yellow should only probably be fixed
[12:55 PM] horowsah: Thanks jflat!
[12:57 PM] jeff101: it's like Foldit has 2 manifolds: the cis one and the trans one, and you need to use special tools to switch manifolds
[12:57 PM] jeff101: some tools will optimize just within one manifold or the other
[12:57 PM] horowsah: yeah, sometimes it's too constrained for the tools to do much
[12:59 PM] horowsah: Thanks all, I have to go to a meeting, but great questions!
[1:00 PM] jeff101: thanks for coming. these meetings are helpful.
[1:00 PM] pc: thanks ^^
[1:01 PM] Nicm25: thank all!

[agcohn821]

[3:07 AM] bkoep: Hello everyone! Welcome to Office Hours!
jeff101: How are things going?
[3:09 AM] bkoep: Things are still a little hectic! Who knew a website had so many moving parts?
[3:09 AM] bkoep: But I think we are steadily quashing all of the issues, and hope to reach equilibrium before the weekend
[3:10 AM] HuubR: In my client, the light next to Notifications is not green but grey, and it says "Attempting to connect to the foldit servers…"
HuubR: Is that a known issue, or just something on my side?
[3:10 AM] jflat06:
Thank you everyone for your patience so far (edited)
[3:10 AM] jflat06: @HuubR that sounds like the IRC issue, which is on our side
[3:11 AM] jflat06: we're working on trying to get that fixed, but there are some more critical issues we need to get to first
[3:11 AM] beta_helix: And thank you again to everyone who has reported bugs!
[3:11 AM] HuubR: I think we should thank you and the rest of the team for your hard work!
[3:11 AM] jeff101: Yes. Change is always hard.
[3:11 AM] alcor29: How are we looking on the crash fix for 2196?
[3:13 AM] jeff101: I think Aunt Dean would pass out virtual cookies if she were here.
[3:13 AM] beta_helix: @alcor29 How are we looking on the crash fix for 2196?
[3:14 AM] bkoep: I believe @Sciren (he/him) has identified the bug and is working on a fix as we speak
[3:14 AM] alcor29: Thx bkoep and Sciren
[3:15 AM]Sciren (he/him): I have indeed! I have tracked down the source and and working on patching it up
[3:16 AM] bkoep: Aside from the 2196 crash, is there anything preventing people from playing Foldit like you were last week?
[3:17 AM] HuubR: Well, I've been struggling with my view presets, but I got that sorted now. See
[3:17 AM] HuubR: https://fold.it/forum/bugs/client-crashes-during-restart-when-show-backbone-issues-is-on#post_73654
[3:18 AM] alcor29: Don't think so. Chat is out, but I imagine that can be fixed soon.
[3:18 AM] jeff101: I figured things would be in flux for a while, so I am still using old clients. I'm doing chat thru an old client that was running before the website upgrade.
[3:19 AM] jeff101: One of my old clients needed a restart, so I am running it offline now.
[3:19 AM] bkoep: @HuubR excellent, thanks for filing a bug report!
[3:19 AM] bkoep: (Also, excellent bug report)
[3:19 AM] HuubR: yw :-)
jeff101: I think Aunt Dean would pass out virtual cookies if she were here.
[3:19 AM] jflat06: Speaking of which, we should make sure she has some version of the candle from the old site
[3:19 AM] HuubR: and thank you :-D
[3:20 AM] rmoretti: Note that at this point any work done on old clients won't be contributing to Foldit science. (Aside from the IRC chat, everything will be stuck on your own machine.)
@jflat06 Speaking of which, we should make sure she has some version of the candle from the old site
[3:20 AM] beta_helix: Yes! Her and the others
[3:20 AM] jeff101: I know, rmoretti, but it raises my knowledge, and that can help Foldit long term.
[3:23 AM] beta_helix: You can always use an old client just to chat, and the new clients to fold
[3:24 AM] jeff101: Maybe later I will take that plunge.
jeff101: I noticed in the 3D viewer on the website for puzzle 2197, there were isolated dots for val30 and gly9. What are those?
[3:32 AM] jeff101: also, my user profile at https://fold.it/users/421719
[3:32 AM] jeff101: https://fold.it/users/421719
[3:32 AM] jflat06: one second!
[3:32 AM] bkoep: The 3D viewer is Mol* (the same viewer used on the PDB website). I believe Mol* "infers" covalent bonds by detecting atoms that are close enough to covalently bond. If two atoms are placed too far apart (by a careless crystallographer, maybe) then Mol* will think those atoms are not bonded.
[3:33 AM] jeff101: used to have an image of Harry Potter's Sorting Hat on it. I don't see that image now.
[3:33 AM] jflat06: currently trying to improve our SSL support, but something is going haywire, so I may wait until things quiet down for the night to try again
[3:34 AM] jeff101: perhaps the pdb for 2197's starting structure is missing some atoms, and this makes the isolated dots
[3:36 AM] bkoep: I hope it's not missing anything important! You'll remember the previous ED puzzle was missing a whole residue, so we were pretty careful to check for missing atoms this time around.
[3:36 AM] beta_helix: Looking at GLY 9: https://fold.it/puzzles/2013458 it might just be the cartoon mode drawing it that way
[3:38 AM] beta_helix: because you can see that HIS 8 is dismembered as well! instead of drawing the polypeptide chain through both sidechains, Mol just drew the cartoon strand in between them
[3:39 AM] jeff101: on the old web site, I often used the html <pre> and </pre> to get a fixed width font
[3:39 AM] jeff101: sometimes I'd use <pre> and </pre> instead, but it acted the same
[3:40 AM] HuubR: https://www.markdownguide.org/basic-syntax/#code
Basic Syntax | Markdown Guide
The Markdown elements outlined in the original design document.
[3:40 AM] jeff101: on this new site, I think the <pre> order no longer works. [3:41 AM] HuubR: Don't know if you can just click the link above, but I think that does what you want. [3:41 AM] jeff101: the first line now shows with a different color and lists <pre> in its output, and the last line now lists </pre> in its output
[3:43 AM] jeff101: thanks HuubR
[3:44 AM] HuubR: Have you tried indenting your text by at least 4 spaces? It says that should do the trick. Have not tried it yet.
[3:45 AM] LociOiling: good they have a section on "escaping backticks"
[3:45 AM] LociOiling: worse than leeches!
[3:45 AM] HuubR: lol
[3:46 AM] bkoep: Yes! @jeff101 for a single line of code I recommend using a single backtick on either side of your code
[3:47 AM] bkoep: For a block of code, you can use three backticks in a row before your block, and three backticks in a row afterwards
[3:47 AM] bkoep: I'm also not sure about indenting code like the guide suggests; I have not tried it either!
[3:48 AM] HuubR: Just did. Three backticks do not work, but four spaces do.
[3:50 AM] HuubR: I love that on-the-fly preview of a post, provided it is responsive :-)
[3:50 AM] jeff101: one really big thing was recipes. It seemed to list public, group, and self-shared recipes all in one place. Was it only listing ones I had access to, or was it listing all of them? also, do you have to logon to the Foldit site to be able to view the source code for the recipes there?
[3:51 AM] LociOiling: the backticks seemed to work for me
[3:51 AM] LociOiling: https://fold.it/forum/discussion/code-block
[3:52 AM] LociOiling: I put them on separate lines before and after the code
[3:52 AM] HuubR: You're right, Loci. Should have added them on both sides.
[3:53 AM] LociOiling: https://www.markdownguide.org/extended-syntax/#fenced-code-blocks
Extended Syntax | Markdown Guide
Advanced features that build on the basic Markdown syntax.
[3:54 AM] LociOiling: oooh, maybe even syntax highlighting
[3:54 AM] bkoep: @jeff101 Yes you will have to log onto the Foldit site to view recipe source code or edit recipes
[3:55 AM] bkoep: You should only be able to see recipes that you have access to (public recipes, group-shares from your group, or self-shares from yourself)
[3:55 AM] HuubR: jeff101, I think just <pre> and </pre> does precisely what you want, without making the text magenta.
[3:58 AM] alcor29: Has anyone tried adding a recipe that is on the web site but has gone from the in client cookbook. I just did because its an important scrirpt but i can't find it anywhere.
[3:59 AM] LociOiling: no syntax highlighting with the three backticks, too bad
@alcor29: Has anyone tried adding a recipe that is on the web site but has gone from the in client cookbook. I just did because its an important scrirpt but i can't find it anywhere.
[4:00 AM] LociOiling: is it a public or group recipe?
[4:02 AM] alcor29: It was one of my authorized ones. I had changed some values and had shared it to myself.
[4:02 AM] LociOiling: time to say thanks and let bkoep, jflat06, rmoretti, and sciren go
[4:02 AM] alcor29: Oh wow that was quick. Thanks folks.
[4:03 AM] Sciren (he/him): Thank all of you!
[4:03 AM] LociOiling: and beta_helix!
[4:03 AM] beta_helix: Thank you all again… and yes, thank you so much to the Foldit Dev Team for all the work they've put into making this happen!
[4:03 AM] jeff101: thanks for having the office hour.
[4:03 AM] alcor29: Hope you all can get some sleep now.
[4:03 AM] beta_helix: We'll talk to you again at tomorrow's OHs

[agcohn821]

[8:02 PM] bkoep: Hello all! Welcome back to Office Hours!
[8:03 PM] bkoep: How are people finding the new website? Any difficulties?
[8:09 PM] beta_helix: (Feel free to let us know if anything is particularly awesome as well
[8:10 PM] alcor29: Looks like not too many people could make it.
[8:11 PM] beta_helix: That's ok… we had one yesterday, and just wanted to try and reach as many different timezones as possible
[8:18 PM] alcor29: I'm still waiting on the 2196 fix. But I'm sure it will come when Sciren:'s done. Otherwise it will take some weeks to digest it all. I've put some observations in Suggestions. but all OK I think. (edited)
[8:19 PM] Sciren: (he/him): Indeed @alcor29 I am trying to wrap up a fix for the 2196 issue as we speak
[8:19 PM] alcor29: No rush Sciren Thanks.
[8:21 PM] Sciren: (he/him) Sure thing! I just want to make certain you all have the best chance to keep doing an awesome job folding!
[8:22 PM] alcor29: Ty
[8:22 PM] alcor29: I'm assuming you all got a chance to catch up on your sleep.
[8:23 PM] Sciren: (he/him) I'm not sure if I know this word. @bkoep do you know what this 'sleep' word refers to?
[8:23 PM] jeff101: Is the Neual NetMutate tool part of puzzle 2196. Will it launch on Sep 9?
[8:24 PM] alcor29: lol
[8:26 PM] bkoep: @jeff101 Yes! We still need to tie up a few loose ends, but we plan to release an update later today that activates Neural Net Mutate
[8:26 PM] jeff101: is puzzle 2196 meant to be its maiden voyage?
[8:27 PM] bkoep: @jeff101, no there will be a new puzzle 2198
[8:28 PM] jeff101: ok @ Sciren: I'm not sure if I know this word. @bkoep do you know what this 'sleep' word refers to?
[8:29 PM] bkoep: Aha yes, I think you mean std::this_thread::sleep_for, which is useful for slowing down Foldit when it goes too fast
[8:29 PM] Sciren: (he/him) Oh yes! I think I have seen some documentation about this.
[8:30 PM] jeff101: I read that the old website will be read-only and at https://old.fold.it/
[8:30 PM] jeff101: will we have to login to access the old site? will one login let us see both the old and new websites? will there be a search tool to search both the old and new websites?
[8:34 PM] bkoep: @jeff101 Good question! Yes you will still be able to log in to the old site and you should be able to search it just like before. I don't think the old site and new site authenticate the same way, so you will have to use your "old" password to log into https://old.fold.it/, even if you change your password on the new website https://fold.it/
[8:34 PM] bkoep: Maybe @jflat06 can confirm that….
[8:34 PM] jeff101: will there be a search tool that searches the source code for all the recipes a player has access to at once? this could help if a player is learning to code and wants to find examples of how to use certain LUA commands
[8:35 PM] alcor29: Will there be added later a way to vote things up or down?
[8:35 PM] jeff101: needing 2 different logins for the old and new websites sounds clunky
[8:36 PM] beta_helix: Great questions, jeff101!
[8:37 PM] jeff101: if you want us to search old Feedbacks so we don't duplicate them, it would help to have one search tool check both the old and new websites
[8:38 PM] rmoretti: All the Feedbacks from the old site should have been ported to the new site in the Bugs or Suggestions forums.
[8:38 PM] alcor29: Can the old feedbacks just be imported to the new discussions or bugs etc?
[8:38 PM] jeff101: are both the old and new websites supposed to contain all the same content, at least when the new website went up yesterday?
[8:38 PM] alcor29: oops crossed lines
[8:39 PM] beta_helix: yes… so if you notice something missing, please let us know!
[8:41 PM] rmoretti: There's some content from the old site which hasn't (yet) been migrated (such as PMs), but the intent is that everything should have been migrated. The old site being kept up is a temporary measure, such that users can point out if we missed the transfer of something important. As beta_helix says, if you see something that's missing from the new site, let us know. The intent is that at some point (when we're sure we got everything) the old site will go away.
[8:41 PM] jeff101: ok. is chat working on new clients now?
[8:42 PM] beta_helix: I have to go teach my 2pm class now… take care everyone, and thanks again for your patience and help. We hope to have you "keep up the great folding" again very soon!
[8:42 PM] LociOilingIRC: only 6 here in vet, so I'm guessing no on the chat thing
[8:43 PM] LociOilingIRC: hey, did screenshots get migrated?
[8:43 PM] LociOilingIRC: for example:
[8:43 PM] LociOilingIRC: https://fold.it/portal/files/chatimg/irc_977284_1662566362.png
[8:43 PM] LociOilingIRC: returns a "not found" error
[8:43 PM] jeff101: thanks beta-helix
[8:44 PM] jeff101: I only see 6 vets in built-in foldit chat on my old client
[8:44 PM] alcor29: I think there may be things missing in the recipe import, but will take weeks or maybe months to know. My recent experience with backups and importing is that files always get lost and some that transfer are damaged. It's never perfect. May have something to do with moving data between two different formats. (edited)
[8:46 PM] alcor29: Thanks beta_helix
[8:46 PM] jeff101: yesterday my new profile page had a link to an image of the Harry Potter Folding Hat, but the link no longer worked. I'd imagine other images I saved to the old site have disappeared too.
[8:47 PM] jflat: @jeff101, the old site should essentially be stuck in time, with only the ability to log in, but not otherwise interact @alcor29: I think there may be things missing in the recipe import, but will take weeks or maybe months to know. My recent experience with backups and importing is that files always get lost and some that transfer are damaged. It's never perfect. May have something to do with moving data between two different formats. (edited)
[8:47 PM] bkoep: It is definitely possible some things are missing from recipe import! We've done our best to detect and report import errors (after you import a cookbook, https://fold.it/ will display a red box with recipes that failed to import). Please let us know if you narrow it down to a specific recipe, for example.
[8:47 PM] jeff101: I can think of many Feedbacks, Forum posts, and even puzzle comments where I included images. I'd imagine some of these will be missing now too.
[8:48 PM] alcor29: Sure will @bkoep
[8:48 PM] LociOilingIRC: new site can be very slow, but I see my profile image is intact
[8:48 PM] rmoretti: But keep in mind that GUI and Lua v1 recipes are no longer supported – those will not be imported to the new site. (You'll have to re-write them for Lua v2.)
[8:48 PM] alcor29: Yes still very laggy on somethings.
[8:49 PM] jflat: i believe that we can get the images working again, it is on my to-do list today
[8:49 PM] LociOilingIRC: ok, thanks
[8:49 PM] LociOilingIRC: old site is asking for an id and password, using old authentication I guess, but it's not happy with my answers
[8:50 PM] jeff101: For the missing Sorting Hat, my new profile page has a link to where the image used to be on the old foldit site. Maybe all I need to do to fix lost images is change part of the address these links use. WOuld it be wise to link to images now stored in the read-only old website? Where are all the images stored within the new website?
[8:51 PM] jflat: sorry, what is live right now at old.fold.it is not the full site. it is still the maintenance page along with some hosted old files that collaborators need access to. We will make an announcement when the old site is available again.
[8:51 PM] LociOilingIRC: ok
[8:52 PM] jflat: we are trying to be careful not to host the full site until we're sure that you can't make changes, because we want to keep the database properly frozen so that we have an accurate depiction of pre-switch.
[8:52 PM] LociOilingIRC: my profile image link starts with https://fold.it/rails/active_storage/disk followed by lots of gibberish,
[8:52 PM] LociOilingIRC: which is the new system I believe
[8:53 PM] alcor29: Will there be added later a way to vote things up or down?
[8:55 PM] jflat: not in the forum. the forum software plugin we're using does not have support for that, unfortunately.
[8:55 PM] alcor29: TX,
[8:56 PM] alcor29: Thanks developers, admins. Have a nice weekend.
[8:56 PM] jeff101: still 4 minutes left
[8:57 PM] jeff101: when I tried the pdb viewer for puzzle 2197, I wondered if it could show the electron density cloud too
[8:58 PM] jeff101: I was able to find a distance between 2 selected residues, but I didn't get the angle (needs 3 residues) or dihedral angle (needs 4 residues) measurement tools to work.
[8:59 PM] bkoep: @jeff101, no, the viewer will not show electron density from the Foldit puzzle
[8:59 PM] jeff101: I also didn't find a button to quit the molecule viewer, so I used Microsoft Edge's left arrow button to exit it
[9:00 PM] bkoep: There are a lot of tools and options in the viewer! I have not spent a lot of time playing around with it, but there is some documentation out there
[9:00 PM] bkoep: The viewer software is called Mol* (or MolStar)
[9:00 PM] bkoep: https://molstar.org/viewer-docs/
[9:01 PM] jeff101: ok. thanks. maybe there's a good place on the foldit site or wiki pages to post that link.
[9:02 PM] bkoep: If you figure out how to use Mol* features, you could open a new topic in the Discussion Forum to share with other Foldit players (and with the Foldit team)
[9:06 PM] jeff101: will you guys be available in chat this weekend? perhaps another office hour in a few days would be useful
[9:07 PM] jeff101: it might even make sense to extend the deadlines for the present puzzles by a few days
[9:09 PM] jflat: I'll be on intermittently
[9:09 PM] jflat: I think it's probably fair to extend puzzle durations, but we'll talk with the team
[9:10 PM] jeff101: I'm out of questions for now. Thanks for having this office hour.

[agcohn821]

Office hour 10/16/22- led by Sciren

[6:01 PM]Sciren: (he/him): Hello everyone! I hope the day is treating you all well 😄
[6:02 PM] spvincent: hi Sciren
[6:03 PM]Sciren: (he/him): For those of you I haven't met with before I am Sciren. I work for the Meiler Lab at Vanderbilt University on Foldit. My work primarily revolves around the integration of Small Molecule Design Tools and various User Interfaces in Foldit. Feel free to ask many any questions.
[6:04 PM]Sciren: (he/him): Hi spvincent!
[6:05 PM] spvincent: Any news on the results of the last round of small molecule design puzzles?
[6:06 PM]Sciren: (he/him): We will have some news to share about it in the near future, but unfortunately none that I can pass along at the moment. I believe it will be worth the wait though.
[6:07 PM] spvincent: ok, tx
[6:10 PM] spvincent: Right now it seems impossible to add heavy atoms (anythinmg bigger than N) without incurring a penalty of sorts, despite these appearing frequently in real life srug molecules. Any plans to address this?
[6:12 PM] Sciren: (he/him): Are you seeing this penalty mostly in the main score, the objectives, or h?
[6:14 PM] Sciren: (he/him): The broad answer to your question though is yes. We would like to have the score reflecting the best choices possible when building drug like molecules, and that will potentially come with some adjustments down stream.
[6:15 PM] spvincent: The objectives mainly which naturally drags down the main score. Sometimes it's possible to add a fluorine without penalty but thats it really: everything else, ingluding fragment groups containing an S or something, never improve the score.
[6:17 PM] spvincent: As things stand , compounds containg lots of tertiary amines seem to score the best.
[6:17 PM] Sciren: (he/him): I see. The objectives are going to be an iterative improvement process. It is our direct line to players in the puzzle, and we want to make sure they are guiding everyone in the best possible way.
[6:17 PM] spvincent: could be just me thoigh
[6:18 PM] jeff101: I could request some features. It would be good if the compound library would say "your compound is in the library" and mark it in a special way in the scroll down menu.
[6:18 PM] Sciren: (he/him): I appreciate the specific examples! It can help us refine things in the future.
[6:19 PM] rmoretti: @ jeff101 That feature should be in devprev at the moment.
[6:19 PM] jeff101: if there is a score that reflects how similar a compound is to the one you send to the library, please list it in the sroll down menu with each compound
[6:19 PM] jeff101: perhaps always have the most similar one at the top.
[6:20 PM] pc: hi
[6:20 PM] alcor: Hi Sciren:. So we pick #1 option in the Compound Library. Now we add some elements which improve the total score greatly. Do we now need another exact match from the Compound Library or is the #1 plus some added stuff a viable new compound for our purposes? Or do we have to do more iterations till in the end we submit a perfect compound match with an improved total score?
[6:21 PM] jeff101: listing a formula for each compound would help too. like NH3 for ammonia, C2H6 for ethane, etc.
[6:21 PM] Sciren: (he/him): I see what you are saying @ jeff101. It would be helpful for players to know just how close their molecule matches to the ones in the library yes?
[6:21 PM] Sciren: (he/him): Hi @ pc !
[6:22 PM] jeff101: yes. how similar each compound in the library is to your submitted compound.
[6:22 PM] spvincent: What does similarity really mean? Shape? Charge distribution? Other things?
[6:23 PM] jeff101: atom composition? bonding?
[6:24 PM] jeff101: sometimes I want to extend a molecule by adding more atoms in the middle of it. perhaps an insert atom feature would help.
[6:25 PM] rmoretti: @ spvincent The version that devprev is using currently for compound library similarity calculation is "ECFP fingerprints" – basically, the fraction of the different types of local substructures (atom neighborhoods) that match in the two compounds being compared.
[6:25 PM] jeff101: it would also be nice to easily chain phosphate groups together, like in ATP or ADP
[6:27 PM] Sciren: (he/him): Hi @ alcor29 for the CACHE competition we will need to submit compounds that are readily available or producible. This is where the compound library comes into play. The best chance of getting a compound that works for the competition will be to ultimately submit one from the library that closely matches the one you have created.
[6:29 PM] alcor: OK. That the means submit a solution which may be lower scoring but better for CACHE. Thanks.
[6:30 PM] rmoretti: For the competition, we'll only be submitting compounds which exactly match the library. – We'll be looking at the other compounds and may use those to pick library compounds, but only the library compounds will actually be submitted for consideration. – So definitely run your optimized compounds through the library again, and see if you can get something close to it.
[6:30 PM] alcor: Thanks moretti.
[6:30 PM] Sciren: (he/him): Thank you for the feature suggestion @ jeff101 I agree that some form of atom insertion would be helpful, and it is something we will be looking into.
[6:31 PM] jeff101: sometimes I wish I could build the ligand by working on several separate fragments and then bringing them all together … maybe a feature like save custom fragment would help …
[6:32 PM] jeff101: like if we made a glycine fragment and then put several copies of it together in a row to make a polyglycine chain
[6:34 PM] Sciren: (he/him): Absolutely jeff101! This type of building is something that we have been discussing and think that it would be helpful to have.
[6:34 PM] spvincent: Tweak ligand: it would be nice to have control over which part of the molecule stays in a fixed position. Maybe a modifier key.
[6:35 PM] jeff101: I agree with spvincent. perhaps something like pins.
[6:35 PM] jeff101: or freeze atoms
[6:35 PM] Sciren: (he/him): Understandable. So something along the lines of freezing or locking part of the molecule in place while you work on another part would be helpful.
[6:36 PM] jeff101: yes
[6:37 PM] jeff101: say we make a library of 25 compounds in one puzzle but can only explore a few while that puzzle runs. could we save the library of 25 and then use it again in the next CACHE puzzle?
[6:40 PM] Sciren: (he/him): I don't believe we are currently have it set up for players to save their libraries in this manner, but I think this is a really interesting idea. So you would want the ability to basically create your own custom library during one puzzle and use on another?
[6:43 PM] jeff101: like me make a ligand on one puzzle and send it to the search compound library tool. that tool finds 25 ligands most like the one we submit. let us save this list of 25 ligands for use in the next puzzle of the series.
[6:44 PM] jeff101: or to help teamwork, let us share these lists with our teammates
[6:45 PM] Sciren: (he/him): Interesting. I can definitely see the benefits of having a feature like this especially in a collaborative sense. This is something that we can look into.
[6:48 PM] jeff101: if each of the millions of compouns in the overall database has a code name, even if it is something like zz10345bca, list these code names within Foldit. this can make it easier to adapt previously successful compounds and to avoid making the same compound repeatedly
[6:51 PM] jeff101: perhaps let us load these componds by name
[6:51 PM] spvincent: I seem to recall that the compounds were going to made by a company called Enamine. They're based in Ukraine: given the situation there are they still operating?
[6:51 PM] Sciren: (he/him): That is a fair point @ jeff101 knowing the names could help prevent the possibility of iterating over compounds that you have previously gone through.
[6:54 PM] jeff101: I think it can be useful to redo certain compounds. Some compounds can bind the protein several ways. Retrying a compound may find its best binding spot.
[6:55 PM] rmoretti: Enamine did have a rough patch early in the conflict. However, they were able to move their operations to locations in Ukraine (& elsewhere in Europe) which were less affected. They're back up and running, and I think they're mostly through the backorders they accumulated when they couldn't make compounds.
[6:58 PM] alcor: Could the 25 or so silmilars that are given us come up with a percentage number so we know how similar they are to our creation?
[6:59 PM] jeff101: it would help if there was a LUA command like GetAtomName(segnum,atnum). It could return things like S P O N C H Cl for the atom #'d atnum on the segment #'d segnum.
[7:01 PM] alcor: "come up with" = display in the little box.
[7:03 PM] Sciren: (he/him): That is a fair point @ alcor29 as @ rmoretti mentioned in the latest devprev they compounds are sorted and displayed based on similarity, but I can see why having additional information displayed would be helpful.
[7:03 PM] jeff101: what is the difference between the protein structure in 2210 vs 2213 ? I think 2210 had PDB#'s like 201 B while 2213 has 201 A instead.
[7:04 PM] Sciren: (he/him): Agreed @ jef101 Adding LUA hooks is on the list of features to implement to help with Small Molecule Design.
[7:06 PM] spvincent: Are chiral centers an issue in this type of puzzle?
[7:08 PM] rmoretti: @ jeff101 They're the same protein, just different crystal structures from the wwPDB. They should be the same sequence, just a (slightly) different backbone structure. The A/B is somewhat arbitrary. (The protein crystalized as a dimer in the asymmetric unit. For the previous puzzles we used the B chain, but for this puzzle we picked the A chain.)
[7:08 PM] jeff101: if you send a chiral compound to the compound library, will the compound library notice it? like if you send it a left-handed one, will it return 25 left-handed ones?
[7:08 PM] jeff101: are they h from the same pdb file?
[7:10 PM] jeff101: I know the helicase can change shape as it moves substrates and products around
[7:14 PM] rmoretti: The chiral information is sent to the compound library, but how much the chiral information is used in the search is a bit of an unknown. I believe it's the case that two molecules which differ in chirality will look more the similar than two which differ in atom identity, so I don't think you'll exclusively get one chirality.
Chirality is important for binding to proteins, but it's not necessarily well controlled by chemical synthesis. There's a fair chance that when we order molecules from the library we won't necessarily get just the one chirality, but they'll come as a racemic mixture of h. That would reduce the effective concentration of the active molecule by half (or quarter, if there's two chiral centers, etc.), but a 2-fold reduction in apparent affinity still will let us see decent binding.
[7:14 PM] rmoretti: The structures aren't from the same PDB, but they are from similar PDBs (from the same publication, just with different molecules bound.)
[7:17 PM] jeff101: thanks for all your answers
[7:18 PM] Sciren: (he/him): I wanted to thank everyone who came out today and asked such amazing questions! Unfortunately we are going to have to bring the office hour to a close. You all have been amazing, and as always I stay impress with your amazing folding(and small molecule) contributions! Happy folding!
[7:19 PM] alcor: Thanks @ Sciren , @ rmoretti . Good talk.

[agcohn821]

Office hour 11/5/22- led by bkoep

[12:00 PM] bkoep: Hello all! Welcome to Foldit Office Hours!
[12:01 PM] jeff101: is the office hour now?
[12:01 PM] jeff101: hi bkoep
[12:01 PM] bkoep: I'll be hanging out here in chat for a bit if you have any questions about recent puzzles, Foldit science, or if you want to talk about protein design!
[12:02 PM] bkoep: Yes, sorry there was short notice for this office hour. We only announced it yesterday
[12:02 PM] HuubR: Hi @ bkoep, jeff101, and others.
[12:03 PM] Nicm25: hello ! some could have seen different dates…. ;)
[12:03 PM] Nicm25: https://fold.it/forum/news/office-hours-nov-5
@ Nicm25hello ! some could have seen different dates…. ;)
[12:04 PM] bkoep: Oh no! That's a bad typo
[12:05 PM] bkoep: I suppose it won't do much good to fix it now…
[12:05 PM] bkoep: Hi @ jeff101, @ HuubR, @ Nicm25 ! How's folding going?
[12:06 PM] HuubR: Focusing on the CACHE Challenge, mainly. That is indeed a challenge!
[12:06 PM] jeff101: fine. the calm between puzzles for me.
[12:06 PM] Nicm25: I'm not currently active due to lack of computing resource allocation, but listening for news.
[12:07 PM] jeff101: I've been cleaning up loose ends from 2219. Not ready to start the next CACHE puzzle. Would help to have feedback about all the rounds we've done so far.
@ Nicm25 I'm not currently active due to lack of computing resource allocation, but listening for news.
[12:07 PM] bkoep: Sorry to hear that, hope to have you back in the saddle soon
[12:08 PM] HuubR: I have a question about uploading designs for Scientist. You ask us in just about every Lab Report to do so, but would it be possible to give some short feedback about the submitted designs? E.g. in the form of a Foldit PM (or Discord DM, if applicable)? That would certainly motivate me to submit designs more often.
[12:08 PM] jeff101: if we've been going the wrong direction so far with CACHE, would be good to know while we still have a few weeks to change direction
[12:10 PM] bkoep: @ jeff101 I haven't been watching the CACHE puzzle results closely (these are @ rmoretti's puzzles), but my understanding is that things are going pretty well—at least for players that are using the Compound Library
[12:13 PM] jeff101: comments to individuals like "we really like your 17944.860 +8500 on puzzle 2219" would help focus efforts. otherwise we have to judge for ourselves.
[12:13 PM] HuubR: Using a compound from the Library is now being encouraged with a 2000 point bonus, in Puzzle 2222 :-)
[12:13 PM] jeff101: like if we made many compounds that bound the protein away from the special hbond sites in 2219, are these compounds not useful?
[12:14 PM] jeff101: would have been nice to know that sooner
[12:14 PM] bkoep: @ HuubR I can understand that, about feedback for Scientist-shares. Unfortunately, we don't have the time to respond personally to each and every Scientist-share. And I wouldn't like to go through PMs, because I think feedback can be helpful for all
[12:16 PM] bkoep: I was hoping that players would find the 3D viewer useful for this. If you open the Load/Save Solutions menu, you can "Copy to Clipboard" a shared solution and then paste a 3D viewer on the website (in the Forums, for example) so others can see and comment.
[12:17 PM] bkoep: If you post your solution like this in a discussion thread, or in the puzzle comments, then tag me (like @ bkoep) and I'd be happy to take a look at it
[12:18 PM] jeff101: does the "Copy to Clipboard" give pdb format that you can paste into a text file?
[12:19 PM] HuubR: Puzzle comments sounds like a logical place. Will you receive an alert even when I would submit a design after closing?
[12:20 PM] bkoep: No, "Copy to Clipboard" just generates an HTML tag that tells the website to render the 3D viewer wherever the tag is pasted. It looks like: < pdb-viewer solution_id="407389202"/>
[12:21 PM] jeff101: I noticed that even in puzzle 2219, you can load a solution as a guide, and it shows the guide's ligand. this is helpful.
[12:21 PM] bkoep: Yes, users will receive alerts for tags anywhere on the website at anytime, regardless of puzzle status! Good question
[12:22 PM] HuubR: :-)
[12:23 PM] jeff101: I was looking at Small Molecule Properties today. one called isosurface is interesting. what does it show?
[12:26 PM] jeff101: is it rendering the surface of the protein that is near the ligand? what do the red gray and blue colors mean?
[12:27 PM] jeff101: should we try to aim ligand O's at the blue parts and ligand N's at the red parts?
[12:27 PM] bkoep: Yes exactly. Isosurface shows you the shape of the binding pocket around the ligand. It can be helpful for visualizing the shape complementarity of your ligand vs the pockt
[12:30 PM] jeff101: the cation pi and pi pi interactions show similar things. they put speres on tyrosine rings near the ligand and similar rings that are part of the ligand.
[12:30 PM] jeff101: the cation pi interactions shows the above plus it puts spheres on arginines near the ligand
[12:31 PM] jeff101: how should we use these spheres? what parts of the ligand's subscore do they affect?
[12:31 PM] bkoep: Yes the red and blue colors just indicate where oxygens and nitrogens are available for H-bonding. The colors themselves are not a perfect guide about what atoms are needed to make a H-bond. Some oxygens have a hydrogen attached (like SER, THR, or TYR residues); these oxygens can both accept and donate H (depending on the orientation) to make an H-bond
[12:34 PM] HuubR: Gee, I never noticed the different colours. Have now increased the intensity for IsoSurface, and that makes them much clearer. Thanks, jeff101 and @ bkoep!
[12:36 PM] Nicm25: perhaps it appears to define our physical material surfaces, viewing them as voxel models, which would be more consistent with electron density model. like ED puzzles, try to solve puzzles with them in opposite directions :)
[12:41 PM] jeff101: In the Open/Share SOlutions Menu, could you add an arrow to share with Both Group and Scientists ? It would save a lot of time for me. I tend to share many solutions with both Group and Scientists.
[12:41 PM] jeff101: Why press 2 buttons when you could press just one?
[12:42 PM] bkoep: The pi-pi interactions are mostly just for visualization, currently. Aromatic rings like on PHE, TYR, and TRP residues have electrons that are concentrated in the pi-orbitals, on either face of the ring. This extra concentration yields a slightly negative charge, which can have a minor attractive effect for positive charges (cations) like ARG or LYS sidechains. These pi orbitals can also stabilize one another somewhat if two aromatic rings are stacked on top of one another; this is the pi-pi interaction. Foldit is not great about modeling pi-interactions, so these things probably won't have a huge effect on your score. But that could change in the future! And the visualizations can be sometimes useful for advanced users or educators
[12:44 PM] bkoep: could you add an arrow to share with Both Group and Scientists
Hmm, we could consider this. But generally we like to keep the UI as simple as possible (already very difficult for a game like Foldit), which usually means displaying the fewest number of buttons possible.
[12:45 PM] jeff101: other combination buttons don't seem needed, but the group+scientists one would help a lot.
[12:46 PM] jeff101: like if you want to share w/group, you don't need to share it w/yourself
[12:46 PM] jeff101: and if you want to share w/scientists, you don't need to also share it w/yourself
[12:48 PM] Nicm25: Suggestion: add options The […] Button , other combinations that less used but necessary.
[12:48 PM] jeff101 it would still help if the compound library would give unique code #'s for each compound.
[12:49 PM] bkoep: @ jeff101, can you elaborate? Maybe can you describe how you would use such a code?
[12:49 PM] jeff101: like a library based on my solution Je1 could overlap some with a library based on my solution Je2
[12:50 PM] jeff101: I already use codes like Je1-3 and Je2-6 to mean the 3rd compound in Je1's library and the 6th compund in Je2's library
[12:50 PM] jeff101: but it is possible that Je1-3 and Je2-6 are identical
[12:52 PM] jeff101: so far I compare their molecular weights and other small molecule properties, but it is possible to have 2 different compounds match on all these things
[12:54 PM] jeff101: if there are 100,000 compounds in the overall library, they could have unique code #'s like 1 to 100,000.
[12:54 PM] bkoep: I see. I'm not sure how the compounds are indexed in the library, but they might have a useful tag like this…
[12:54 PM] bkoep: Yes the compound library is pretty big (more like 100,000,000) so a numeric ID would probably have to be pretty long
[12:54 PM] jeff101 35,030 might have different stereochemistry from 35,031 and 35,032, but otherwise all 3 compounds are the same
[12:56 PM] jeff101 you could use a mix of letters and #'s to give more codes. 10 #'s + 26 letters gives 36^n different codes if n is the # of digits.
[12:56 PM] jeff101 I mean if n is the # of characters
[12:58 PM] HuubR: There is a simple check that you can do: when you have Je2-6 as your active solution, and you load the library for Je1, it will show the nearest match at the top. If Je2-6 is in library Je1, you will either see it as "Exact match" or "Similarity 1.000000", depending on stereochemistry.
[12:59 PM] bkoep: Yes, I think it will depend on the library (the library is maintained by a separate group unrelated to Foldit)
[12:59 PM] jeff101: you already use codes w/many characters for each new version of cmp-binary, cmp-resources, and cmp-database
[12:59 PM] Nicm25: simply displaying the 'rational formula' is sufficient?
[1:00 PM] HuubR: I think Loci mentioned the other day that there is a SMILES description string in one of the log files, when you submit a compound to the Library. Now I understand that there is no unique way to derive a SMILES string from a structure: even a simple molecule like ethanol can be described starting at either end. So you could still have different strings for the same compound.
[1:01 PM] jeff101: is 'rational formula' like C2H3O2 or H2O ?
[1:01 PM] Nicm25: that's it
[1:02 PM] jeff101: I did see some SMILES format in a log file, but I think most of these were for many repeats of the same very long error message
[1:04 PM] jeff101: I didn't see flagrant lines like "accepting compound SMILES format
[1:04 PM] bkoep: @ HuubR Actually, I believe SMILES has some clever conventions for how atoms are ordered and which atom goes at the beginning of the string. So you can (for the most part) unambiguously translate from SMILES to structure and back again
[1:04 PM] bkoep: Although at the same time I seem to remember there are also some rare edge cases where this is not strictly true
[1:11 PM] bkoep: Alright all, I have to run now. It was great chatting!
[1:12 PM] HuubR: Thanks for your time, @ bkoep, and have a nice day, all!
[1:12 PM] bkoep: Remember to tag me with "@ bkoep" if you post any 3D viewer solutions on the Foldit website, so I can drop in and take a look (edited)
[1:12 PM] Nicm25: many thanks for talking
[1:13 PM] HuubR: I think you meant "@ bkoep"?
[1:13 PM] HuubR: :-)
[1:13 PM] bkoep: (whoops)
[1:14 PM] jeff101: thanks for coming

[agcohn821]

Office Hour 12/15/22- led by beta_helix:
[11:01 AM] beta_helix: Hello everyone, we're about to start today's Foldit Office Hour
[11:03 AM] beta_helix: I'm beta_helix, I'm an Associate Professor (http://www.cis.umassd.edu/~fkhatib/cv.html) who has been working on the Foldit project since 2008…
[11:05 AM] MicElephant: I have a question about scoring: The current revisiting puzzle has 50 segments. Each segment scores with less than 100. But in the puzzle I have 10.000 points. How is this calculated?
[11:06 AM] MicElephant: I don't find such info on the website.
[11:06 AM] agcohn821: I'm also here–my name is Ariana and I'm the Community Manager here at Foldit
[11:06 AM] beta_helix: I wanted to reflect on the past 12-18 months of Foldit (which has changed a lot compared to 2008 ) and would love to hear your thoughts on the recent changes (AlphaFold, Neural Net Mutate, he new website) and where Foldit is heading next! (edited)
[11:09 AM] beta_helix: @ MicElephant the score function is essentially a sum of many different scoring components, you can see these pre-residue scores by pressing Tab on a residue (this is where you saw the segment scores, right?) but there are other scoring components in the Rosetta energy function, which is what Foldit uses as its scoring function. (edited)
[11:12 AM] spvincent: Are there plans to synthesize some of the results from the desihn puzzles?
[11:12 AM] MicElephant: thanks, but its difficult to optimize somthing, when you don't know how it works. Perhaps you can put alink on the website to further details
[11:14 AM] beta_helix: @ spvincent Yes indeed! The new Neural Net Objective (which awards a bonus for great AlphaFold predictions: https://fold.it/forum/news/lab-report-39-boost-your-score-with-alphafold) will hopefully result in even more synthesizable designs!
[11:16 AM] beta_helix: As all those who have been folding design puzzles for the past decade, they are especially tricky to synthesize correctly (which is why over the years Foldit has added so many different tools/scoring mechanisms to try to ensure that the top-scoring designs will fold).
[11:16 AM] jeff101: https://foldit.fandom.com/wiki/Special:Search?query=foldit+subscores&scope=internal&contentType=&ns%5B0%5D=0&ns%5B1%5D=2900
Foldit Wiki
Special:Search

[11:17 AM] spvincent: The neural net objective is a bit of an "all or nothing" thing. Maybe it should be more graduated.
[11:17 AM] jeff101: you can search the Foldit wiki for topics like "Foldit subscores"
[11:17 AM] beta_helix: Thanks jeff101 Gotta love the FolditWiki
[11:18 AM] MicElephant: @ jeff101: thank you for the link
[11:18 AM] jeff101: https://foldit.fandom.com/wiki/REF2015
[11:18 AM] beta_helix: We are excited about the new tools that have come out in the past year to improve your designs! The thing we want to avoid the most is all of you working very hard on a top-scoring model, only for it to have 0% chance of folding because Foldit is missing a biochemical/physics component in the scoring function.
[11:18 AM] jeff101: the one above will give you more info than you want
[11:20 AM] beta_helix: @ spvincent I will bring that up at today's Foldit Team meeting. Thanks! It is tricky when an objective is so binary… but for some things, that's really the case. Clashes for example are "all or nothing", as opposed to voids, right?
[11:22 AM] jeff101: the Foldit Newsletters are helpful too: https://fold.it/forum/discussion/foldit-newsletters
[11:22 AM] beta_helix: Yes, the newletters are an awesome source of info for new players and veterans as well!
[11:23 AM] spvincent: Are clashes all or nothing? In that case why is the Clashing subscore graduated.
[11:25 AM] beta_helix: I like to think of it as all or nothing (atoms should never be clashing with one another) but the score is graduated based on the distance. But if 2 atoms occupy the exact same spot, that is just impossible. (But I was using that example in terms of Foldit visualizations: you never want to see clashes, but you can't always get rid of every void)
[11:27 AM] spvincent: tx
[11:27 AM] beta_helix: But you are correct that energy functions do not like to optimize for binary values (and neither do Foldit players!)… so I'll see what the Foldit devs can do to improve that.
[11:28 AM]beta_helix: How do you all feel about the new website?
[11:28 AM] beta_helix: It only took us 14 years to update it
[11:30 AM] jeff101: change is hard.
[11:31 AM] spvincent: I'm afraid I rather preferred the old one. But then one gets set in ones ways.
[11:31 AM] beta_helix: This is why I wanted to ask! What is missing from the new one?
[11:32 AM] jeff101: the other day I wanted to share a recipe with group, but it seemed like I only had 2 choices: share it with myself or share it with everyone.
[11:32 AM] beta_helix: Yes, I saw that post, we are discussing that today as well.
[11:33 AM] MicElephant: the old and new site both have there advantages/disadvantages. E.g. tables with only 5 lines makes it difficult to serach for ones results a yaer ago. The old site had 25 lines
[11:33 AM] beta_helix: Thanks for pointing that out, as this is obviously functionality that we need to add back in!
[11:33 AM] spvincent: I think there'a forum thread on peoples grievances but for me one thing that sticks out is that there is too much whitespace
[11:33 AM] jeff101: a flagrant link to the Foldit wiki would help
[11:33 AM]beta_helix: Have any of you tried the protein viewer? https://dev.fold.it/puzzles/2013451
[11:33 AM] jeff101: a more obvious way to search the new website would help
[11:34 AM] beta_helix: I agree with that as well!
[11:34 AM] beta_helix: @ MicElephant that is a similar point when searching
[11:35 AM] beta_helix: These are all very helpful, please keep 'em coming!
[11:35 AM] beta_helix: The reason I ask about the protein viewer is that this is precisely the type of tool that we could not implement with the old website…
[11:36 AM] beta_helix: and if we want to incorporate cool new features in the future, the old drupal POS website wasn't going to cut it.
[11:36 AM] beta_helix: (but we obviously still want to keep the functionality of the old website, especially searching and linking to the wiki!)
[11:37 AM] jeff101: the protein viewer is nice. I think it is very like the one built in to the pdb site.
[11:37 AM] beta_helix: Yep
[11:37 AM] beta_helix: we totally stole it from there
[11:37 AM] beta_helix: I mean, "were inspired by"
[11:37 AM] jeff101: can it show electron density clouds?
[11:38 AM] beta_helix: Good question! I would hope so… at least a static cloud, I don't think you could adjust it like in Foldit (but maybe!)
[11:39 AM] beta_helix: I'll ask about that… I have never loaded a cloud on the PDB viewer, but that doesn't mean it doesn't exist
[11:40 AM] jeff101: it might help, because often I don't want to stop a Foldit cleint's recipe to show its ED cloud
[11:40 AM] beta_helix: That is a good point.
[11:42 AM] beta_helix: Improving ED is one of the new features we hope to roll out in 2023
[11:42 AM] jeff101: are we giving you good results in the ED reconstruction puzzles?
[11:42 AM] spvincent: I know there are only so many resources that can be put into Foldit development but I'd like to see a greater emphasis on fixing existing bugs.
[11:44 AM] beta_helix: @ jeff101 Yes indeed. Our hope is to eventually let you improve your ED cloud based on your own model, so you'd be able to reload in a better ED cloud, rather than waiting for us to post a Round 2 puzzle with a better cloud that was generated from the best solutions.
[11:44 AM] MicElephant: @ spvincent: that's true
[11:44 AM] beta_helix: @ spvincent well that was the wrong time for me to post that last paragraph!
[11:45 AM] beta_helix: The dev team is working on existing bugs, but perhaps our priority queue doesn't match the player needs?
[11:45 AM] beta_helix: Are there some obvious bugs that have been unresolved for too long that we need to address?
[11:47 AM] jeff101: I sure with an old feature would return. I liked to be able to Tab on one residue and then use the arrows to scan along the protein to see neighboring residues' subscores too
[11:47 AM] MicElephant: the trim tool lets my clients always chrash. Ihave put a log file on the website
[11:47 AM] beta_helix: Yes, the trim tool is actively being fixed thanks to your feedbacks
[11:47 AM] spvincent: :). I suspect every player has a different pet peeve but I would like to see the screenshot issue fixed. I must say though that Foldit has become a lot more stable over the last few months which is good.
[11:48 AM] beta_helix: Screenshots and Tab scanning… great.
[11:48 AM] beta_helix: I will raise those up the queue today. And most importantly: if these are not solvable with the new site, we will let you know (rather than giving you false hope that we will someday fix it)
[11:49 AM] MicElephant: the undo window sometimes doesn't work when you are manually folding. Not sure when and why
[11:50 AM] beta_helix: Hmm… that's tough, because we really struggle when it's not reproducible. Which is why the screenshot and Tab bugs are easy to diagnose
[11:50 AM] MicElephant: I will raise a bug, when I know more
[11:51 AM] jeff101: Mic … maybe you are doing a long wiggle and want undo to only go back part way
[11:51 AM] jeff101: it is good to stop wiggle at regular intervals so there are more spots you can undo to
[11:52 AM] MicElephant: no, after starting the puzzle, wiggle, shake, wiggle, and the undo stays empty
[11:53 AM] jeff101: can you tell us how we did in the CACHE challenge? On their schedule it looks like Foldit had to send them our best 100 ligands by Nov 27.
[11:54 AM] jeff101: which ligands did you choose?
[11:54 AM] beta_helix: I knew someone was going to ask that It's like CASP, in that we won't know the results until the organizers announce them… at which point we'll have a full blogpost about it
[11:55 AM] beta_helix: Sadly, I have no idea… sorry. Sciren, you on?
[11:55 AM] agcohn821: @ Sciren (he/him)
[11:55 AM] beta_helix: thanks Ariana (edited)
[11:57 AM] beta_helix: They will be at today's Foldit meeting and I will ask them if they can post which ligands they selected… as well as when we can expect the results to be announced.
[11:58 AM] beta_helix: I am not at all familiar with CACHE, but do know the CASP15 meeting just ended: https://predictioncenter.org/casp15/doc/CASP15_Meeting_Program.pdf
[11:59 AM] beta_helix: Any last comments/questions?
[12:00 PM] spvincent: I'm slightly surprised CASP is still around after the success of AlphaFold2 and TrRosetta
[12:00 PM]beta_helix: If you know a better time for Office Hours, please let us know for next month!
[12:00 PM]beta_helix: @ spvincent They also do RNA now
[12:01 PM]agcohn821:
Yes please–office hours feedback is welcome!
[12:01 PM] beta_helix: and Deep Learning doesn't really work on RNA because there aren't enough solved structures yet.
[12:01 PM] agcohn821: Thank you all for your incredibly helpful feedback and your contributions–we appreciate you all so much!
[12:01 PM] beta_helix: If you're interested, here are the CASP15 results: https://predictioncenter.org/casp15/index.cgi
[12:01 PM] MicElephant: thanks for your answers. Good luck, when fixing the bugs!
[12:02 PM] beta_helix: You can click on the different Rankings to see the tables/graphs
MicElephant: thanks for your answers. Good luck, when fixing the bugs!
[12:02 PM] beta_helix: Thank you! We'll need the good luck
[12:02 PM] spvincent: tx beta_helix
[12:02 PM] beta_helix: Thank you all!
[12:03 PM] beta_helix: Keep up the great folding and I'll talk to you next year

[agcohn821]

Office hour 2/19/23- led by jflat06

[12:01 PM] jflat06: Hey all, office hours is starting now. I am jflat06, a software developer on the Foldit team. If you have questions related to the Foldit software or server, I'll do my best to answer them.
[12:02 PM] pc: hi
[12:03 PM] pc: we can do a "here" maybe to notify users ^^
[12:03 PM] jflat06: True!
[12:03 PM] jflat06: @here Office hours starting now!
[12:03 PM] joanna_h: Never managed to be here for one of these :D
[12:04 PM] pc: Do you know informations about Neural Net Mutate ? I use it, it is nice, but have some limitations.
[12:05 PM] joanna_h: Do you have any plans to take advantage of GPU acceleration?
[12:05 PM] pc: for exemple, it is good to find core hydrophobic AA, but it ofen give alanine for most of the segments. (edited)
[12:05 PM] pc: maybe there is a problem in implementation ?
[12:06 PM] joanna_h: Another question, do you have any plans for people to run local alphafolds (if they have the hardware.) Would be nice to run it more often and get faster analysis.
[12:08 PM] pc: when NN Mutate give a AA : we don't realy know if it is a good one, or if our folding is bad, and he give us a random one.
[12:08 PM] joanna_h: Got one more, currently the alphafold show interface just gives some colour indication of where it is bad. It also can be pretty useless when you are nearly there. Is there a possibility of getting alphafold to recommend what should be there (AA wise) to improve the fold or similarity?
[12:08 PM] MicElephant: jflat is now in the global chat
[12:08 PM] joanna_h: oh
[12:09 PM] jflat06: Sorry about that! We're back
[12:10 PM] jflat06: minor technical hiccup
[12:10 PM] pc: welcome back ^^
[12:10 PM] joanna_h: oh lol @ pc Do you know informations about Neural Net Mutate ? I use it, it is nice, but have some limitations.
[12:12 PM] jflat06: I have some basic knowledge of it. A lot of ML algorithms aren't very transparent on why they do things, and I can see that being frustrating. It's not perfect but statistically, it has shown to provide good results, though.
[12:12 PM] pc: maybe Neural Net Mutate can give information in what AA is good or bad, like Neural Net with AlphaFold ? (edited)
[12:12 PM] pc: or here is more segment folding.
[12:12 PM] jflat06: That might be possible - I am not sure what extra information NN Mutate provides.
[12:13 PM] jflat06: The NN Mutate algorithm is called MPNN if you want to look it up - I'm pretty sure something has been published by now.
[12:14 PM] jflat06: "Joanna_H: Do you have any plans to take advantage of GPU acceleration?" In Foldit, probably not for the foreseeable future. I am actually working only part time on Foldit at this point, and my new project is on GPU acceleration of our biochemistry software. So for the future, who knows!
[12:14 PM] pc: Neural Net mutate is usefull for AA in core that I have difficulties to find to have a solid starting protein. But it help have a 90% confidence faster than before ^^
[12:15 PM] pc: off course, I need combine it by hand folding and normal mutate (edited)
[12:16 PM] joanna_h: hopefully sooner than later :D
[12:16 PM] jflat06: "Joanna_H: Another question, do you have any plans for people to run local alphafolds (if they have the hardware.) Would be nice to run it more often and get faster analysis." Not currently. It should be theoretically possible. But there is a lot of very messy code on the backend to interface the alphafold server to the client. Another option might be opening up our PDB loading/saving so that people can run whatever external tools they want in the future, but we aren't there yet.
[12:17 PM] spvincent: It would be nice to see AF for the sandbox puzzle
[12:17 PM] joanna_h: oh, opening up the PDB would be wonderful. If we could import/export that would be amazing.
[12:18 PM] jflat06: "Is there a possibility of getting alphafold to recommend what should be there (AA wise) to improve the fold or similarity?" - alphafold is a backbone prediction algorithm based on sequence. It sounds like you want Neural Net Mutate, which will modify the AA sequence to match what it thinks should be at that location.
[12:18 PM] joanna_h: yes
[12:18 PM] joanna_h: a blend of the two, interacting together to tell you is really wrong.
[12:19 PM] joanna_h: what is really wrong.
[12:19 PM] joanna_h: Sometimes it can be very hard to work out.
[12:19 PM] joanna_h: especially when you are close.
[12:20 PM] pc: there is multiple little features to improve Foldit in navigation, selection.. I posted some suggestion in forum, but I havent get any answer.. But maybe they where read ? I can do more feedbacks, but I dont want to give too mush ^^
[12:20 PM] joanna_h: I'm still thinking about all the possibiliies if we could export/import PDB.
[12:20 PM] joanna_h: oh wow
[12:20 PM] pc: I can give feedback in how I play and how I use tools. Is dev or scientist interested ? Maybe a community manager can answer some question or feedback like this everyweek ?
[12:21 PM] joanna_h: so many secondary tools out there we could use.
[12:22 PM] HuubR: I think importing and exporting would open up a lot of possibilities (beyond my imagination :-), but what about the implications for the competition element? I would not want to use the word "cheating" here…
[12:22 PM] joanna_h: how can you cheat, a lot of do reasearch anyway with things like JPred.
[12:23 PM] pc: For the protein preview 3d in forum, it is a cool feature. But it can be better with an option to show bonds, and maybe void or buns (for binder puzzles)
[12:23 PM] jflat06: @ pc Our dev resources are a little smaller than they used to be, so we aren't able to address as many player feedbacks about things recently. A lot of the dev work is being done by specific teams, like the devs working specifically on the small molecule drug design puzzles.
[12:24 PM] jeff101: are the 502 and such errors all sorted out?
[12:24 PM] pc: thanks ^^
[12:24 PM] jflat06: We do usually glance over feedbacks, at least, but we're probably just bad at responding.
[12:24 PM] joanna_h: @ Jeff101 I would guess not I got one earlier uploading an alphafold.
[12:25 PM] jflat06: @ jeff101 They aren't all solved, but they at least seem to have improved since the fix from a while ago. You can tell me, though - have things been better recently?
[12:25 PM] spvincent: On the small molecule design puzzles. is there any need to share designs with scientists? Or will such solutions be recorded on the server in the batural course of gameplay?
[12:25 PM] guineapig: why is the start point of the frozen part in round 2 different from round one. there are a lot of sidechains which are in another positions. How can you compare the solutions?
[12:25 PM] joanna_h: @ jflat06 it's a lot better, not every 5-10mins but they do still happen.
[12:26 PM] jflat06: "HuubR: I think importing and exporting would open up a lot of possibilities (beyond my imagination :-), but what about the implications for the competition element? I would not want to use the word "cheating" here…" A great question! And probably the reason we haven't just went ahead and done it. We are warming up to the idea, though [12:26 PM] HuubR: :-)
[12:27 PM] joanna_h: I know it's a game, but is not the real point to create good science?
[12:27 PM] pc: In ligand binder puzzles, I have a feedback : I load AF prediction, and I try to keep the pocket for the ligand (bkoep talked about this in last lab report video). But it mostly keep this pocket only in more than AF 88% similarity. Maybe we can give a bonus for players that reach the 88% similarity ? I saw lot of top score solutions that fail at keep the pocket.
[12:27 PM] HuubR: Agree!
[12:27 PM] jflat06: "spvincent> On the small molecule design puzzles. is there any need to share designs with scientists? Or will such solutions be recorded on the server in the batural course of gameplay?" - we always log solutions automatically, but sharing with scientists is a great way of highlighting a solution you care about, especially if it doesn't have a top score.
[12:28 PM] jflat06: Joanna_H> I know it's a game, but is not the real point to create good science? - It's a mix. The competitive element of the game is part of what drives motivation, so you can't ignore it completely. But at the end of the day, a lot of our players are here to help science.
[12:29 PM] joanna_h: Agree, but you can be competative and do good science. And more tools allow for better sceience and more competition D
[12:30 PM] pc: On the small molecule design puzzles, I have some ligand with full objetive but no compound library exact match (but near !). It is interesting to share with scientists ? Sometime it is only one atom, it is likely synthetisable, it makes more bonds, but it is not possible to find a perfect similar ligand in compound library.
[12:31 PM] joanna_h: The old CPU faster/slower argument has always been there. I experienced it, I was on an amd 3400 over Xmas and it was horrible. tick, tick, tick for every change. Now I'm back home on the 5800X with the 3080 it's like watching butter flow down the screen.
[12:32 PM] joanna_h: So in some we all "cheat" deepending on our hardware.
@ pc On the small molecule design puzzles, I have some ligand with full objetive but no compound library exact match (but near !). It is interesting to share with scientists ? Sometime it is only one atom, it is likely synthetisable, it makes more bonds, but it is not possible to find a perfect similar ligand in compound library.
[12:32 PM] jflat06: That is probably a good question to post on the forum - the puzzle specific questions are better answered by people who actually know the subject matter. I am not a very good biochemist!
[12:32 PM] MicElephant: What I am currently missing is to get a feedback on the puzzles with severall rounds. E.g. we played the 7th round of the FMN binder, but do not get any hints what shall be changed from previous solutions.
[12:35 PM] pc: there is a long time we haven't any feedback if there is lab test in binder puzzles. Last was there is one year ago. I know that scientists always watch solutions, but is there a hope that some lab tests will be done for all thoses proteins we made last year ? Or maybe some where already done ?
[12:36 PM] jflat06: MicElephant - Yeah, I can understand it is difficult to know how to proceed without feedback on your solution. Unfortunately, it is quite a bit of work for bkoep to give even the feedback he does, and there is only 1 of him! My suggestion would be to use the website solution posting feature and to solicit advice from other players. Some of our players are getting pretty good at design, so they can likely spot problems even if the team is busy.
[12:38 PM] joanna_h: Do you think there will ever be any possibility of accessing external code outside of LUA? Effectively we create native code that runs a lot faster and the data is passed down and returned?
[12:38 PM] jflat06: We tend to be a little tight lipped about lab testing, because it's a very slow process with lots of incremental preliminary results that don't mean much in isolation.
[12:39 PM] MicElephant: A question about 2266: will loading saved solutions work in near future on that puzzle?
[12:39 PM] spvincent: I knpw it's a lot of work to expermentally create and test these proteins. But it would still be interesting to know which ones are being made and what the results are. Even if they don't fold properly or don't bind as expected, etc.
[12:40 PM] jflat06: Joanna - Possibly. There are efforts to open source Rosetta, the software that Foldit is based on. So it's possible that in the future Foldit could be open sourced as well and players could design features and integrate them into the game.
[12:40 PM] joanna_h: I think people might be noticing my questions are all about external intergration and computation outsourcing.
[12:41 PM] jflat06: As a question from me, since I am curious - do people feel as though the 502 / server problems have gotten better over the past several weeks?
[12:41 PM] joanna_h: they have
[12:41 PM] joanna_h: I've had 1 all day
[12:42 PM] jeff101: do you ever e-mail players to say "We really like your B42 design on puzzle 2258, please try to make more like that in the future" ?
[12:43 PM] guineapig: i cannot use the private message feature at the moment.
[12:43 PM] jflat06: I think the only time we email players directly is concerning paper authorship. Otherwise, I think the feedback is usually limited to the public website posts.
[12:44 PM] jeff101: or "all these strange designs you're making in 2273 aren't helping us much. we like what you did in puzzle 2257 better. please make more like you did there."
[12:44 PM] jflat06: Hmm… I just tested it and it works for me. Is that your user account name, guineapig?
[12:45 PM] guineapig: yes. I am not the only player with that problem
[12:46 PM] spvincent: I haven't had any server problem recently
[12:46 PM] jeff101: one thing I attribute to the 502 errors is when I start a certain recipe, it pauses a long time before it brings up its input menu. This didn't used to happen.
[12:46 PM] jflat06: The only reason I can think that it wouldn't work is that you aren't approved on the website, but I just checked and you are, so I'm not sure what's up.
[12:46 PM] spvincent: And I second jeff101's comments on feedback!
[12:46 PM] MicElephant: I can answer to private messges, but not start a new one.
[12:46 PM] pc: sometimes I had to restart foldit, but was not a so big problem for me ^^
[12:46 PM] joanna_h: Thank you @ jflat06 for answering my questions, I love the fact that you are considering opening up more interefaces for external interaction. Unfortunately I now have to go. Again, thank you for being here to answer the questions.
[12:47 PM] jflat06: Thanks for coming!

[agcohn821]

959 AM

horowsah

Hi all- I'm here for the office hour; will be hanging out for an hour and hy to chat and answer questions. I'm not very familiar with the ligand design series; I do mostly the reconstruction puzzles, both the original sort and the refine density.

horowsah

If you're curious, I also have a wet lab where we look at how RNA affects protein folding and neurodegeneration, so that's another topic I like to jabber on about

beta_helix

When are we gonna get RNA into Foldit for real?!

horowsah

Would love to have had that hen 0 years ago

horowsah

Anyways, there was one thing I wanted to share for sure.

Anfinsen_slept_here Am I right to worry about model bias in the refine densitypuzzles? Any plans toshow us Rfree? If that's not too wonky of a question

A

horowsah

Oh good question, I'll answer that first

horowsah

Yes, model bias is an issue

horowsah

Not so much in the old school reconstruction puzzles based on the way I prep the maps

horowsah

But in the refine density puzzles it's almost surely a problem.

horowsah

We want to integrate Rfree, but we have some bugs in it that have been tricky

horowsah

So for now model bias will limit us in those puzzles. That said, it's still helping in a bunch of cases, so it's one of those things where we think improvement is still wortwhile even while more improvement is still likely later

t

Anfinsen_slept_here Thank you. Do you wanna say more about where the ED maps come from

horowsah

You never can predict when an improved structure could really help someone out in their research.

horowsah

Ok, now onto the sharing bit

horowsah

If you go to https//pdb-redo.eu/

horowsah

there's a text box to search for pdb entries

horowsah

Thanks!

beta_helix

https//fold.it/forum/discussion/how-to-access-the-foldit-models-on-the-pdb-redo-site#post_79688

horowsah

This is the complete way to find foldit stuff

beta_helix

It's not super easy to access, but you can follow the above directions

beta_helix

to see the Foldit models that improved on the original PDB and on the PDB-REDO solutions

beta_helix

PDB-REDO is an automated server that tries to improve all models submitted to the PDB

beta_helix

(but even that isn't always enough)
beta_helix pinned a message to this channel. See all pinned messages.
—
/28/25,

horowsah

the thing I love most is that when you go to the actual page for the entry on PDB-REDO now, when a Foldit player improved upon it and had the top structure, the Foldit player gets credit right at the top of the page

horowsah

For example

LociOiling

nice

beta_helix

https//pdb-redo.eu/db/04l

horowsah

So as we keep running reconstruction puzzles and Foldit players find issues in these structures, more and more will start to look like this

horowsah

Anyway- wanted to share that and also get people's thoughts on it

t

Anfinsen_slept_here will we do pdb structures w ligands

horowsah

Ah yes, ligands.

horowsah

At some point, we'll get to that, the issue is primarily how labor intensive it is to set those puzzles up when the ligands change every time

horowsah

But that should be getting easier at some point and then it will be a big thing to do

beta_helix

We know how good you all are at ligands, though!

horowsah

In most structures, the ligands are the most suspect part, so we definitely want to have that as an option in the puzzles

LociOiling

do we know how many PDB-redo entries mention Foldit?

beta_helix

65 right now

LociOiling

not bad

beta_helix

Especially when we haven't had hundreds of puzzles!

LociOiling

yes

beta_helix

@horowsah want to answer this question? "Anfinsen_slept_here Thank you. Do you wanna say more about where the ED maps come from" (edited)

horowsah

Sure!

horowsah

The maps come from two sources- either x-ray crystallography or cryo-EM

horowsah

the prep and downstream analysis of each looks pretty different, but the main part that hens in puzzle is pretty much the same

horowsah

For the reconstruction series, we're using the publicly deposited data and structures in the PDB. I search for those that seem like they could really use some help.

LociOiling

what do you look for in terms of "needs help"?

t

Anfinsen_slept_here When we save a model in a Refine Density puzzle, are we also saving phases?

horowsah

@LociOiling I tend to look first for geometry and clash issues, as Foldit players generally are really good at helping with those. Issues of fit to the maps often also improve, but not always by as much. I'm now starting to look for those also that PDB-REDO itself is unable to fix, to see if Foldit players can help those cases where human intervention really seems needed

t

Anfinsen_slept_here when we reload a model, do we get the new map, or does the ED revert/

horowsah

On the saving phases question- we actually aren't if I recall correctly. Instead, we always recalculate them.

beta_helix

Andreas will know

Andreas

we have the 'unphased' density based on crystallography data saved at all times, and phases are always calculated from the model in the player's solution at the time or running the refine density tool

t

Anfinsen_slept_here btw, i love the ease and power ofthe refine density function

beta_helix

We've wanted that in the game since 202!

beta_helix

We just needed to find Andreas to pull it off!

Andreas

oh and to finish answering the model saving part of the question, the model that is used to calculate the density is now saved as a part of a solution

Andreas

which may or may not be the same as the current model itself

ButterKnights

a "best refine density" button would be nice. similar to how credit best work but it is only the scores post pressing the "refine density" i have noticed that progressing off of this yield better scores then off of "Credit Best"

Andreas

do you mean a button that would load the 'best' density back into the game?

ButterKnights

are the highest scores from the refine density puzzles actually good folds scientifically?

Andreas

if so the issue with this would be the fact that there really isn't a best density in terms of Foldit scoring. models are scored based on how well they match the density, so better density is more implicitly determined based on the overlap between the model matching the density and the model being good

ButterKnights

@Andreas yes but specifically only the scores immediately after pressing the refine density button. these scores may be worst then the "Credit best" solutions

horowsah

@ButterKnights Short answer on scientific usefulness is- in many cases yes. The limiting factorbig issue is that without the Rfree stat, overfitting becomes an issue in some cases.

Andreas

the top solutions by r free are almost always in the top 3-5 solutions by Foldit score

Andreas

and in many cases are the top Foldit solution

beta_helix

If you remember from the Foldit cryo-EM paper (and the x-ray too), we had to look at a ton of metrics to determine which model was "the best"

beta_helix

beta_helix

beta_helix
Click to see attachment

beta_helix

http//www.cis.umassd.edu/~fkhatib/Papers/NatureComm.pdf

Andreas

@ButterKnights i'm not totally following but one thing to keep in mind is that rephasing doesn't always make the density better. there's a chance that when you rephase it's making the density worse if something in your model isn't correct. furthermore, even with 'good' density you will likely still need to tweak the model after refining to better match this new density
beta_helix
Click to see attachment

beta_helix

http//www.cis.umassd.edu/~fkhatib/Papers/PLOSbio209.pdf

t

Anfinsen_slept_here Its hard to know when refinement is 'done"

beta_helix

Very true!

horowsah

I believe a very well known crystallographer once said, "Refinement is never completed, merely abandoned"

t

Anfinsen_slept_here yep

beta_helix

We also have to realize the limits of crystallography (though it's best not to think about that too much when it's our gold standard ) (edited)

LociOiling

any insight on when refine density is likely to help? it's a little unpredictable to me…sometimes early is good, but often it's later on in the puzzle run

horowsah

Great question

horowsah

In theory, the closer the solution gets to correct, the less refining the density will help

horowsah

But it's not linear

horowsah

So sometimes something that you think might be a minor change could help a lot, but a major change might help less

horowsah

It doesn't follow human logic perfectly

LociOiling

that has been our experience

horowsah

That comes from it actually dealing with the fourier transform of the data, which I don't think humans can intuitively understand

beta_helix

Yep… I don't get it

alcor29

How big a factor do you think serendipity helps?

horowsah

In my experience, an unfortunately large amount

Andreas

'luck is where preparation meets opportunity'

horowsah

I tend to think of it like you're digging for gold one shovelful at a time

horowsah

And you have very little indication of where it is before you start digging

horowsah

When you find it, often times it leads to more, but not always. But finding separate pockets can be really hard.

beta_helix

We should rebrand the game as digging for gold!

alcor29

"Chance favors the prepared mind"

ButterKnights

why would transposing (no relative atom movement) a whole fold with the exact same shape give a different fold it score after refine density.

t

BletchleyPark good morning to you, have my questions from the website already been discussed ?

beta_helix

Not yet… it's been busy and we were hoping you'd show up

LociOiling

(I would add ButterKnights is also known as bravosk8terboy on Foldit)

t

BletchleyPark thank you, please also post the links in the chat in that website post

beta_helix

https//fold.it/forum/news/office-hour–28-2025#post_79685

horowsah

@ButterKnights If I understand the question right, so the foldit score is based on the current fit to density, so when you refine density, the foldit score changes

t

BletchleyPark links from the chat

Horowsah

But there is actually a bit of play in it, so it might be numerically slightly different from time to time

t

BletchleyPark referring to pdb entries, papers etc.

beta_helix

https//fold.it/forum/discussion/how-to-access-the-foldit-models-on-the-pdb-redo-site#post_79688

beta_helix

http//www.cis.umassd.edu/~fkhatib/Papers/NatureComm.pdf

beta_helix

http//www.cis.umassd.edu/~fkhatib/Papers/PLOSbio209.pdf

horowsah

On the opennes question- this is part of when we started writing, we reached out to a number of players for strategies they use on these puzzles

beta_helix

(those are all the links so far, I believe)

horowsah

they're now all in the supplement of the preprint

beta_helix

I'll answer jeff0's question towards the end (in case he is able to make it)

horowsah

There's a great variety in there, and so that will hopefully help everyone improve
@horowsah
@ButterKnights If I understand the question right, so the foldit score is based on the current fit to density, so when you refine density, the foldit score changes

ButterKnights

the exact same shape in two different locations gives a different score post refine density. are you saying that the fit to density can change even if the fold does not change?

beta_helix

Preprint https//www.biorxiv.org/content/0.0/2024.06.9.599674v2.full
Reconstructing Biological Molecules with Help from Video Gamers
Foldit is a citizen science video game in which players tackle a variety of complex biochemistry puzzles. Here, we describe a new series of puzzles in which Foldit players improve the accuracy of the public repository of experimental protein structure models, the Protein Data Bank (PDB). Analyzing the results of these puzzles showed that the Fol…

beta_helix

"Supplemental Information-Foldit Player ED strategies" is at the bottom of that page/PDF

horowsah

@ButterKnights I think I didn't fully understand the question the first time- there is a positional dependence to the whole protein as the electron density is essentially fixed in space. However, the density also has a repetitive nature to it based on the underlying crystal shape. So it won't be the same for every puzzle

horowsah

On the absolute rotation and positioning as scriptable functions, I agree that these would be good to have. We'll have to think on the best way to implement those.

horowsah

And on the question of when we get those broken puzzles with crazy scoring, we have long conversations on these

horowsah

On the one hand, it isn't scientifically useful when it hens and we don't want to reward it

horowsah

On the other, if people put time into it without knowing what is going to hen, they could be in for a shock
new

horowsah

Ideally, we catch them quickly, but that doesn't always hen
new

BletchleyPark Would some of those structures have any relevance in real life ?
new

horowsah

Some still will
new

horowsah

just the scoring is broken
new

horowsah

But others, not so much

BletchleyPark thank you for answering my questions

horowsah

No problem!

beta_helix

Shall we answer Jeff0's I'm curious how we did on Puzzle 254 (https//fold.it/puzzles/203979) Unsolved Voltage-gated Ion Channel Cryo-EM Density.

horowsah

hmmm

beta_helix

Our collaborators plan to repost it

horowsah

I'm trying to figure out which things Diego didn't answer yet

alcor29

@beta_helix It would be good feedback to know who the 65 are.

beta_helix

I think it's just a general question about the results of the puzzle

horowsah

but generally speaking, it looked really good to me. The issue is that there was one thing that was in the top structure that doesn't jive with some biochemical data, suggesting we'll likely have to tweak the puzzle a bit and probably let people have another go at it

beta_helix

If you click on the pinned link I posted, you can see all 65

alcor29

Thanks

horowsah

But reposting takes some decision making and we don't want to do it wrong

beta_helix

When we publish the paper, we'll have a nice table for everything… but this server will constantly be growing with more and more answer Foldit solutions

Andreas

we’ll have some graphs to share progress on density puzzle results soon!

Andreas

specifically comparing refinement and non refinement puzzles

beta_helix

Jeff0, wherever you are, we plan to repost that puzzle soon. These are close collaborators at UC Davis from the Rosetta Community. You can hopefully see that by how they answered questions in the puzzle comments https//fold.it/forum/puzzle-comments/uc-davis#post_7948

BletchleyPark "When we publish the paper ", is it already known where it will be published ?

ButterKnights

will we ever get a atom type function in LUA? ligand puzzles are tough to identify what atoms are what by eye.

beta_helix

@ButterKnights I will bring that up to the Drug Design te!

new

horowsah

So the paper is in an interesting spot. After we posted the pre-print, it was seen by organizers of the CCP4 study weekend conference, and they asked us to submit it as part of a special issue to Acta Cryst D, which is a very well regarded crystallography and structural biology journal. So we'll be submitting it there soon.

new

beta_helix

Any last questions for today's Office Hour?

horowsah

It's not that usual to get those sorts of requests from really good journals based on pre-prints, so it's another mark in how people really value the work of the players of Foldit

Anfinsen_slept_here Are there times when winning and non-winning structures are only different by tiny RMSD's? My loser self was wondering?

horowsah

That's a good question that I'm not sure the answer to

horowsah

Presumably sometimes yes

beta_helix

You were so close, Anfinsen!!!

Andreas

i think i can answer this actually

beta_helix

It actually goes back to those graphs I showed earlier about what is "the winning" structure

Andreas

if solutions are super close in terms of RMSD they are usually in the se “cluster” of solutions, only one of which is sent for further refinement and analysis via pdb redo

Andreas

the solution selected for each cluster is always the top by Foldit score

BletchleyPark Shouldnt the whole cluster then be credited ?

beta_helix

The PDB-REDO server can't handle us sending it 00,000+ Foldit solutions, so we cluster them here first. By cluster, we don't mean smoosh them together… but group the ones that look "similiar enough" to one another, and only pick "the best" from that group

Andreas

if a solution is slightly worse by score but determined to be too similar in terms of RMSD to analyze separately then it is still empirically worse
new

beta_helix

You can think of it as selecting a representative model from each structural topology
new

horowsah

I have to go, but this was great, thank you all!
new

BletchleyPark Would that then be the absolute top scoring solution in all cases ?
new

alcor29

Thanks all. Good meeting. Can we have them more often please?
new

beta_helix

I want to thank you all so much for chatting with us today!
new

beta_helix

Yes @alcor29 we should do this more often

Anfinsen_slept_here Yes REALLY

beta_helix

Does this time of the workday work for most of you?

BletchleyPark often

beta_helix

We know how hard it is to find a timezone that works for everyone

BletchleyPark Thanks for your time, and the participant's time
@
BletchleyPark Would that then be the absolute top scoring solution in all cases ?

beta_helix

So the absolute top scoring solution will always be "the best" of its cluster

BletchleyPark but there may be other clusters with other best solutions, I see.

beta_helix

but what if the 269th solution looks completely different and just has a slightly lower score

exactly

BletchleyPark thank you for clarifying

beta_helix

We'll make sure to post this transcript ASAP

beta_helix

of course! Thank you all for coming… keep up the great folding!

BletchleyPark thanks, have a great day !