Developer Chat

@betahelix Hello everyone! We'll be starting our Scientist Chat in a couple minutes... 12:58
bkoep Hi everyone! 13:01
Wbertro yes, 1659b needs babysitting, cannot not leave it unattended without crashing 13:02
Wbertro log.txt never reports any helpful info except UNHANDLED EXCEPTION 13:03
Wbertro so cannot play it very much as I am doing something else 13:03
Wbertro sorry for the side track 13:04
@betahelix No worries at all. 13:04
Wbertro I'll be quiet now 13:04
bkoep @Wbertro, please, share details about your crashes in a Feedback on the website! 13:05
bkoep jeff101 has started one here: 13:05
bkoep 13:05
Wbertro it is just like I said nothing in the log.txt file, nothing much I can say above that, it crashed doing all different things 13:06
Wbertro except for a bunch of errors in the log.txt 13:06
Wbertro I'll post that 13:06
@betahelix Thanks Wbertro 13:07
@betahelix Let's get started by answering the questions in the comments: 13:08
@betahelix 01010011111 asked about the exciting results of Google's AI DeepMind: 13:09
Wbertro Posting Done 13:09
@betahelix The RosettaCommons actually invited the heads of DeepMind to the annual Rosetta Developers meeting, as keynote speakers... 13:10
@betahelix but they are unfortunately unable to make it this year. So there are definitely plans for cooperation in the future, but sadly not as soon as the Rosetta community would have hoped. 13:11
bkoep Also, before I forget, we have a new member of the Foldit team to introduce, sgorji 13:13
sgorjiIRC Hi! 13:13
spvincent hi 13:14
robgee hi sgorji 13:14
@Susume2 welcome sgorji 13:14
sgorjiIRC I'm a new student spending some time in the Baker lab, getting to see all the cool behind the scene action in Fold it 13:14
@betahelix Welcome to your first Foldit chat (of hopefully many!).   13:14
sgorjiIRC Thanks for all that you guys do! 13:15
sgorjiIRC This is such a cool project 13:15
bkoep Next up, @robgee asked about the aflatoxin puzzles from last year: 13:16
bkoep "Is there anything you can say about aflatoxin: Have the foldit results produced anything interesting? Are the foldit results of any value in any way? Are you brewing up results in the lab ?" 13:16
bkoep Unfortunately, we haven't had any updates from the Siegel lab to share 13:17
bkoep I understand they are still carrying out experiments based on some of the Foldit player designs 13:17
bkoep But at one point they were having trouble with the enzyme activity assay (to detect breakdown of aflatoxin), so they may still be sorting all of that out 13:18
@betahelix We will check in with them for an update. 13:19
robgee Do Mars and Thermofisher share their results with you guys 13:20
bkoep @robgee, Mars and Thermofisher are only providing support for the research (some funding and materials for experiments), but they're not actually generating any results 13:21
bkoep All of the experiments are being carried out in the Siegel lab at UC Davis, who are usually very good about sharing data when they have it 13:21
robgee Oh i see, thanks. 13:23
bkoep Next, from @Susume: 13:25
bkoep "In 2018 the Baker lab had two publications about the first successful denovo designs of beta barrel proteins using Rosetta. Foldit players have also submitted a variety of beta barrel designs: have any foldit barrels been successful in Rosetta testing?" 13:25
bkoep Yes 13:25
bkoep Sorry, that was early 13:26
bkoep We haven't tested any Foldit beta barrels in the lab 13:26
spvincent have you run them through Rosetta? 13:26
bkoep I meant to say "Yes, recently some of our colleagues in the Baker lab had some very cool results with designed beta barrels" 13:26
bkoep Unfortunately, beta barrels are still very difficult to design. 13:27
bkoep For testing Foldit designs, we still rely heavily on Rosetta@home folding funnels to decide which designs to test. And so far, no Foldit-designed beta-barrels have passed our Rosetta@home test. 13:27
bkoep There’s some more info about Rosetta@home in this old blogpost: 13:27
bkoep 13:27
@betahelix How about any beta helices? (sorry, I couldn't resist) 13:27
bkoep Of course, the Rosetta@home test is imperfect, so there’s a good chance that perfectly good designs are slipping through the cracks. 13:28
bkoep Eventually, we’d like to carry out high-throughput experiments, so that we can test thousands of Foldit designs at once. Then we can be less picky about the designs we choose to test in the lab. 13:28
bkoep @betahelix, in fact we DID attempt to test a beta helix design in the lab 13:29
bkoep Unfortunately, though, the protein failed to express in E. coli, so we don't know if it might have folded up or not 13:29
bkoep That result will be a part of the Foldit design paper that will be published soon 13:29
@betahelix speaking of which... 13:30
@betahelix We should answer any questions about the upcoming Foldit Design Paper! 13:30
@Susume2 On the couple of barrel designs that i ran thru Rosetta on a lark, Ginzu did a rather poor job of selecting templates, so Rosetta didn't really stand a chance 13:31
@jeff101 how about the electron density puzzles from Fall 2018? 13:32
@betahelix Yes, we will talk about those for sure today! :-) 13:33
spvincent In the paper, you have a list of proteins but there doesn't seem a way of going from those descriptions to the original FoldIt structures 13:33
bkoep @Susume, yes I think Ginzu relies on sequence homology so if your design sequence does not resemble any natural protein it might have problems 13:33
@Susume2 It would be nice if you could feed Rosetta@home a template chosen by the designer rather than Ginzu 13:34
bkoep @spvincent, yes, when the paper is published, it will include the protein models for all of the tested Foldit designs (as downloadable Supplementary Information) 13:35
@jeff101 I'm guessing any server that uses contact maps to work out structures will have trouble with our designs. Aren't the contact maps based on databases of natural proteins? 13:36
bkoep @Susume, the Rosetta@home protocol doesn't actually rely so much on domain recognition software like Ginzu. However it DOES try to use the protein sequence to pick short (9-residue) fragments of known protein structures, and then assemble those fragments into decoy structures. 13:37
rmorettiIRC @jef1101 Yes, most of the contact prediction methods used rely on multiple sequence alignments of closely related (naturally occurring) proteins. 13:37
bkoep @Susume, we would like to add a tool to Foldit so that players can see whether the selected fragments are a close match to the design 13:38
@Susume2 that would be cool 13:39
@betahelix Did anyone have any other questions about the Foldit Design paper? 13:42
@jeff101 do you have cover art? 13:42
@jeff101 whose art did you pick? 13:42
@betahelix Great question! We were able to send more than just one. 13:42
@Susume2 do we have a publication date? 13:43
@betahelix Even though in the end it will be Nature's decision... if we're lucky enough to get one picked. 13:43
@jeff101 it will be in Nature, right? what issue #? 13:43
@betahelix Indeed it'll be in Nature :-) I remember that we submitted some cover art suggestions for the first Foldit Nature paper, but sadly they didn't pick one (although we did get a caption on the cover!) 13:44
robgee I tried to make some art but Gimp has a steeper learning curve than  foldit ;) 13:44
@betahelix We do not know the issue number yet or the publication date. 13:44
@betahelix @robgee I hear you! 13:45
bkoep We also had a question from fiendish_ghoul about symmetric design in Foldit: 13:45
bkoep In symmetry puzzles there is a core existence filter ensuring that each monomer unit has an individual core. But there are some structures in PDB in which monomer units lack their own core or have only a small part involved in it (for example 5Z1O, 4S37 and 4OSD). How do such proteins fold and is it feasible to try to design such structures in foldit? 13:45
@betahelix We will of course post that information as soon as we know it, but we know that we have to be patient... as the final steps always take a while with journals. 13:45
Marvelz robgee best look at some youtube tutorials for that it helps a lot. Also using inkscape to vector your gimp images is awesome. 13:46
bkoep There are indeed some natural proteins that assemble into well-folded symmetric assemblies, even though the individual chains do not form well-folded domains. 13:46
bkoep However, we still do not have a very good understanding of how these proteins fold, and that sounds to me like an extremely difficult design problem. 13:47
bkoep For now, I think there is enough to be done by designing symmetric assemblies out of well-folded building blocks 13:49
bkoep We understand these systems much better, and it's still a very difficult problem. 13:49
@betahelix Any last design questions? 13:49
bkoep In fact, previous Baker lab members have had considerable difficulty designing symmetric assemblies out of proteins with beta sheets (most Baker lab symmetric designs are completely helical) 13:50
spvincent I believe there was aquestion about the restrictions on symmetry in theese 2d puzzles 13:50
spvincent sorry: symmetric dimer not 2d 13:51
bkoep We think this is an area where Foldit players might be able to outshine scientists 13:51
bkoep @spvincent, yes, we haven't forgotten about Skippysk8s's question! I think we can answer that one offline, though. 13:53
spvincent ok 13:53
@betahelix @jeff101 asked about the ED puzzles from last Fall... 13:53
horowsah I can give an update there 13:53
horowsah Those ED puzzles were from a cryo-em structure, and we compared the solutions of the Foldit players to that of the original microscopist who solved it as well as a few algorithms 13:54
horowsah and perhaps not surprisingly, the best Foldit models were better than the other sources 13:54
horowsah better than that from the microscopist, even 13:55
@betahelix I certainly was not surprised one bit! 13:55
horowsah we're still in the process of analyzing the data, but that much seems clear now 13:55
@betahelix For all 4 of the puzzles we posted (imagine if we had been able to give you more!) 13:55
horowsah we think that this is going to be very interesting to the microscopy community 13:57
@betahelix Another interesting result: In each of the 4 puzzles, the very best model was the top-scoring Foldit prediction! 13:57
horowsah yes, which was not necessarily the case with older EM puzzles 13:57
horowsah sorry, ED puzzles 13:57
horowsah this is probably due to the differences between maps generated by crystallography vs microscopy 13:57
@betahelix (so we at least seem to have gotten the score function pretty spot on when it comes to cryo-EM puzzles!) 13:58
horowsah I would expect that this is also an area where Foldit players could have a big impact in the future as a result 13:58
@betahelix So congrats to all of you first this, we are very excited by this result! 13:58
@betahelix We will be trying to get our hands on more cryo-EM puzzles very soon. 13:59
@betahelix I did have a question about size (which has always been Foldit's Achilles's heel). It seemed that the 231 residues puzzle wasn't as terrible as we on the Foldit Team feared it might be... how was it? 14:00
@jeff101 I don't recall. I enjoyed all the EM puzzles I tried this fall. 14:01
@betahelix That 14:01
@betahelix 's a good sign if you don't remember any particular one being painful. 14:01
@jeff101 you should look at comments folks made in Feedback or the puzzle comments 14:02
@betahelix We are curious if the cryo-EM puzzle size isn't as big of a deal as it was with crystallographic density puzzles. 14:02
@betahelix 14:02
@jeff101 or follow Vet & Global Chat as it happens ... does anyone on the Foldit Central read the Chat logs regularly? 14:03
@betahelix That was good news when we read it, I was just curious if anyone else had any night terrors about it :- 14:03
@Susume2 giving longer to work on a big puzzle is key 14:04
@betahelix Yeah... and 2 weeks is generally enough for the larger ED ones, right? 14:04
@Susume2 takes multiple sittings for the initial hand work, and scripts take longer to run 14:04
@Susume2 2 weeks is good IMO 14:06
@jeff101 I've suggested many tools to make ED/EM puzzles easier. One would be to display a calculated ED based on the present structure. Then we can compare it with the actual ED. 14:06
horowsah in many ways, that's what the analysis post-Foldit does 14:07
horowsah it's something that typically crystallographers will look at before they build and after they build, but not while they build 14:07
horowsah but it can be a nice way to verify that you're on the right track 14:08
@jeff101 make it a view option like cartoon or trace tube are already 14:08
@jeff101 or another option for the ED menu where spacefill, wireframe, mesh, etc. are listed 14:09
@jeff101 if it takes too long to calculate, the player can turn it off & save time 14:09
@jeff101 or for spot checks, turn it on, wait a bit, compare to the experiment's ED, then turn it off 14:10
horowsah the one thing I will mention is that with crystallography maps, I expect that feature will be more useful than with EM maps, due to the sources of error in crystallography that aren't apparent in EM 14:10
horowsah but it's a good idea 14:10
@jeff101 also let the user translate the protein from one unit cell to another 14:11
@LociOilingIRC I like the idea of having calculated density be an option on the ED menu 14:11
@jeff101 to help find the unit cell where the protein overlaps best with the displayed ED cloud 14:11
@LociOilingIRC separate RGB/alpha settings for both densities 14:11
@jeff101 yes loci I agree 14:12
@betahelix I think we need a Dev Chat about ED! 14:12
robgee The big ED puzzles ran ok for me, bit slow, not too crashy. 14:12
@Susume2 give us an ED puzzle as a reminder first ;) even if a revisit 14:13
@betahelix I'd like to refine a crystal ED map based on the current Foldit model, then you could load in a better map! 14:13
@betahelix Will do! 14:13
bkoep I think our hour is about up. But before we go, I do have a brief personal announcement. 14:13
bkoep I recently defended my PhD thesis, titled “Protein design by citizen scientists”! 14:13
bkoep Obviously, Foldit players were essential to all of the work I presented. So I want to extend a personal Thank You to all of the Foldit players who have participated in Foldit over the last six years. 14:14
@jeff101 congratulations! 14:14
@MikeCassidytoo Congrads!! 14:14
NinjaGreg Yes, nice! 14:14
@Susume2 *\o/*  Congrats!! 14:14
@LociOilingIRC Dr. Bkoep 14:14
robgee Go Dr.bkoep 14:14
Marvelz Congrats! 14:14
NinjaGreg Will you send us a link to your thesis? 14:15
@jeff101 will you continue to support Foldit in your next position? what will you do next? 14:15
bkoep Thanks, all! I do consider it a shared achievement. 14:16
bkoep We really did learn a lot about protein design, and I hope everyone is excited that the work will finally be published for everyone to read about! 14:16
bkoep I'm actually not sure how the thesis will be published (electronically only, of course). But I will see about sharing it when I know more. 14:17
@jeff101 UMI / Dissertation Abstracts used to do it ... 14:18
bkoep @jeff101, I am still figuring out my next plans. For the time being, I'll be continuing to work with Foldit from the Baker lab. 14:18
@jeff101 they'd make microfiche, print copies, and microfilm copies that people could buy 14:18
@betahelix I say you tweet out your entire thesis! 14:19
@jeff101 :) 14:19
@betahelix Any final questions before we wrap up today's chat? 14:19
@jeff101 thanks for having it 14:19
@betahelix Thank you all for coming! (and everyone else from reading the transcript!) 14:20
@jeff101 are these new design puzzles like 1659b collabs w/an external group 14:20
@betahelix But most of all: thank you all for the amazing work that you do. You never cease to amaze us, and the entire protein folding field! 14:21
bkoep Yes, thanks for a great chat! 14:21

Generated by 2.17.0 by Marius Gedminas
- find it at!

Download links:
  Windows    OSX    Linux  
(10.12 or later)

Are you new to Foldit? Click here.

Are you a student? Click here.

Are you an educator? Click here.
Other Games: Mozak
Only search
Recommend Foldit
User login
Top New Users

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons