Developer Chat

smortier Welcome to today’s Science Chat!   13:12
smortier We are going to mostly focus on the questions that were posted by players in the SciChat announcement, but if you want clarification or have a quick, easy-to-answer question, feel free to ask and we’ll try to address it over this next hour. 13:12
bkoep Hi everyone! 13:13
jflat06IRC Hello! 13:13
smortier Let's dive right in. 13:13
smortier Our first question: Puzzle 1387b had a very strange molecule at segment 30. I'm not even sure if it qualifies as ligand, since it's spliced into the backbone just like one of the 20 usual suspects. Among the odd features of segment 30 was a "shy" hydrogen that seemed to have an unlikely position. What does segment 30 represent?   13:13
bkoep Segment 30 from that puzzle is a sugar-based chemical group called a glycan. In biology, glycans are often added to the surface of a protein for any of a number of different purposes. 13:14
bkoep In particular, the body uses glycans for recognition of “self” and “non-self” proteins. Influenza has evolved to coat some of its surface proteins in glycans, as a strategy to disguise them as human proteins, and to evade the human immune system. 13:14
bkoep The position of the the “shy” hydrogen was erroneous, but probably shouldn’t affect the puzzle results too much. 13:15
bkoep In general, we don't want to bind glycans, since they are all over the place in the body, but we want players to be able to see it on the target protein so they can avoid it 13:15
smortier QUESTION: Regarding 1387b are there any updates on whether the designs foldit players made seem to fold as designed in the lab, and if so, do they seem to bind to the hemagglutinin?   13:16
@TimovdL Then binding there should be penalized maybe 13:16
bkoep @Timo, yes, if that seems to be a problem in many designs, we would want to penalize interactions with that residue 13:17
bkoep Unfortunately, we haven’t been able to test any of those designs yet. 13:19
bkoep There were also some problems using the regular design filters on that puzzle. We’re working to improve how these binder puzzles are set up in Foldit, so expect more binder design puzzles in the future! 13:19
smortier QUESTION: Is there any way to check whether the predictions made by foldit players are close to the native fold or not? If so, what are the results?   13:20
Susume2 (that was in ref to dysferlin linker domain puzzle) 13:21
bkoep I think this question is in reference to the dysferlin puzzles (like puzzle 1390)... 13:21
bkoep Until we have a structure of these dysferlin domains, we can’t really know if the Foldit models are correct. 13:21
bkoep We have some collaborators who are planning more experiments with dysferlin; their experiments may corroborate or reject the Foldit models, but we won’t know anything about this for some time. 13:21
bkoep I'd also like to say, the Jain Foundation was really excited and appreciative of Foldit players and their enthusiasm for the dysferlin puzzles 13:23
smortier QUESTION: How come the contact map on this puzzle seemed so "chaotic/random" and how come it differs quite a lot from the contact map predicted by raptorx  (link below) 13:23
bkoep If they have additional ideas for Foldit puzzles, I think they will probably come back! 13:23
smortier   http://raptorx.uchicago.edu/ContactMap/myjobs/83006813_233351/   13:23
bkoep Ultimately, contact predictions are based on sequence alignments, with sequences from many different related proteins 13:24
bkoep If we are unable to find a large number of quality sequence alignments, then the predicted contacts will be very “noisy,” and probably won’t contain very much actual information. 13:24
bkoep Any difference in the RaptorX and the Gremlin predictions is most likely due to differences in the sequence alignments used by each. 13:24
bkoep RaptorX doesn’t seem to report much info about the sequence alignments, but you can see the raw Gremlin output here: 13:25
bkoep http://openseq.org/sub.php?id=1480265603 13:25
bkoep In particular, I notice that the “Seq/Len” value is a little low, indicating that Gremlin was unable to find a large number of aligned sequences for this protein 13:25
bkoep RaptorX may have used a different alignment strategy that produced a different set of alignments 13:25
BletchleyPark Could these kinds links please be added in a prominent place on the webpage ? 13:26
BletchleyPark kinds of 13:26
@TimovdL Maybe repost it with the RaptorX contacts? 13:26
smortier That is definitely something I will take to our next team meeting! 13:27
bkoep @Bletchley, yes we could post these links if players find them useful 13:28
bkoep I suspect a lot of the technical details are a bit too far "in the weeds" for most players, and we like for Foldit to feel approachable for novices and new players 13:28
bkoep @Timo, yes we might try reposting the dysferlin puzzle with the RaptorX contacts 13:28
smortier Next question is more of a development question, maybe jflat can chime in here: 13:29
smortier QUESTION: Are there any plans to refine the camera and dragging controls?   13:29
jflat06IRC I like the idea of giving additional options to refine the camera controls - though the full question gave some additional suggestions that I thought I would address. 13:31
jflat06IRC The first was the idea of using WASD+Mouse look to control the camera, as is common in FPS games. 13:31
jflat06IRC While I can appreciate this would be intuitive for people coming from those kind of games, many of our players don't have a FPS (first person shooter) background, so those controls may not provide too much utility to them. 13:32
jflat06IRC An additional point that was brought up is that things such as band-end placement can be very awkward with the current tools. In particular, the question brought up issues about "snapping" unintentionally to the protein backbone when you're just trying to band to space. 13:33
jflat06IRC I'll mention here that holding down the Alt key while dragging a band will temporarily disable snapping :) 13:34
jflat06IRC Aside from that, I can appreciate it is awkward to place a band at the correct depth when viewing it on a 2D monitor. 13:34
frood2IRC WOuld it be a good idea to have a tutortial that explains this stuff to newbs (most vets know these things already) 13:35
jflat06IRC The usual approach to this is to place the band end, rotate 90 degrees, then place it again. This isn't the most efficient way of doing it, but without proper depth perception, doing anything better is difficult. We could perhaps have a camera control for snap rotating 90 degrees? 13:36
@TimovdL On camera controls in ED puzzles, it would be nice to have a special interface to manipulate forground/background, using just the keys in not very handy 13:36
smortier @frood2IRC - that's a really good idea. 13:37
jmbrownlee333 yes. i would love more control of the slab in ED 13:37
jflat06IRC @Timo - you mean in order to trim the density? 13:37
@TimovdL Density and visibility of protein parts 13:38
jflat06IRC I see - so hide anything that is too close or too far from the camera? 13:38
frood2IRC I ubderstand the issues regarding ED - but I honestly believe we have many viewing tools - they work. But many do not understand them. That’s just my tuppence worth  - clearer training on existing facilities again. 13:39
@TimovdL Or making them easier to use 13:39
frood2IRC we can already hide far/near 13:39
frood2IRC sorry - maybe I misunderstand…please carry on TVDL - sorry 13:40
@TimovdL I would like to see something like far near for foreground background 13:41
jflat06IRC We're definitely interested in improving your control over the density (something we've talked about) 13:41
jflat06IRC I think you can probably expect to see some more work done on this. 13:42
smortier QUESTION: Are we less needed for unpredicted proteins? In a paper from 2011, players found the structure of an HIV protein. We don't seem to repeat this success. Are we less accurate with our recipes? or don't you need us anymore for such kind of work? 13:42
bkoep Good question! 13:43
bkoep The biggest reason we haven’t tackled similar problems is that we rarely encounter data like we had for the MPMV puzzle. 13:43
bkoep We need collaborators to supply us with unsolved x-ray diffraction data; x-ray crystallographers are often reluctant to share unpublished data, for fear of being “scooped” by other researchers. 13:43
bkoep Also, automated programs have gotten a lot better in recent years, so there are fewer of these unsolved datasets around. 13:43
bkoep In addition to that, we’ve recently been pushing much harder in the direction of protein design 13:44
bkoep I think there is a greater potential in this area (i.e. protein design) for Foldit players to make more significant contributions. 13:45
smortier QUESTION: Are the Drug Design puzzles coming? 13:46
rmoretti Hi! I'm taking over the DrugDesign project from free_radical, now that he's finished his postdoc and moved on to other things. (Though he'll be finishing up his work by posting puzzles occasionally.) 13:46
rmoretti (I've been 'behind the scenes' working on code development for drug design all this time.) 13:46
rmoretti We've actually just posted a drug design puzzle to devprev - “[DEVPREV] Ligand Drug design test bugs” 13:47
rmoretti We're anticipating releasing more drug design puzzles (including ones to main) shortly. 13:47
LociOilingIRC quick question on that -- will the "experimental" update group be used any more? 13:48
frood2IRC WIll these drug design puzzles count toward rank scoring?  Many are not happy that it should - they feel it is not why they play foldit. 13:48
rmoretti Not generally, at least not for drug design stuff - most of the improvements will likely be in devprev. 13:49
@TimovdL And another one, any special LUA functions for those kind of puzzles? 13:49
jflat06IRC @Loci - yes, though this update group isn't just for drug design. 13:49
LociOilingIRC thanks 13:49
jflat06IRC You can expect us to use it for changes to the program that we want to test in isolation. 13:49
rmoretti We currently don't have any LUA functions enabled for drug design, though that will be something we enable eventually once we get the general kinks worked out. 13:49
smortier Thanks rmoretti for the update! 13:50
rmoretti (That is, we don't have LUA functions specific to drug design yet.) 13:50
smortier QUESTION: Wondering if the loop, helix, sheet, and filter limitations might actually be introducing a “bias”.... and if expanding the library and changing limitations of the structure might actually produce more meaningful data. (after an initial period of relative chaos from a database standpoint)   13:51
bkoep I'm not entirely sure I know what this question is getting at, but I'll try to answer as best I can 13:52
bkoep It is certainly not our intention to limit or bias Foldit players any more than necessary. 13:52
bkoep We’ve curated the available loops to only those that are exceptionally common, in an effort to improve design quality (i.e. the probability that a high-scoring design will fold successfully). 13:53
bkoep It’s an unfortunate side-effect that this limited selection of loops also limits variability in designs, but I think it's a necessary evil (for the time being, anyway). 13:53
bkoep More abstractly: 13:53
bkoep If we had infinite resources, and could test every Foldit player design, then you’re right—the increased variability in designs would contain richer data about protein folding. 13:53
bkoep However, this would be a very inefficient approach to increase our understanding of protein folding and design. 13:54
bkoep With limited resources, we have to think about what experiments will be most valuable to the field. 13:54
smortier QUESTION: Ideal SS sheets are not straight. However, working with initial straight sheets seem to give more points at the end. And tweak too tends to straight the sheets. Are ideal sheets straight or not?   13:56
bkoep There could be a number of reasons that initially straight sheets might lead you to better-scoring solutions, but they are likely artifacts of the game 13:57
bkoep The majority of sheets observed in native proteins have some degree of twist 13:58
bkoep The twist allows the strands to form stronger hydrogen bonds with better geometry 13:59
smortier QUESTION: Why doesn't Blueprint propose us a loop of 3 residues for connecting sheet-sheet?   13:59
bkoep Oddly enough, we simply don't observe very of these loops in nature 14:00
@TimovdL Also with an odd number of segments in a sheet, the blueprint folds the 2 other sheets to the same side 14:01
bkoep It is far more common—at least in highly-stable, well-defined structures—to find 2-residue hairpins connecting sheets 14:01
smortier QUESTION: Have there been any novel-shaped foldit designs (shapes not found in nature) that have performed well in Rosetta or in the lab? 14:02
bkoep As far as I know, the Foldit design crystal structure we discussed in the blog (and in Puzzle 1384) represents a fold not yet observed in nature—though I haven’t yet conducted a rigorous search of the PDB. 14:04
bkoep There definitely exist larger natural proteins that contain that arrangement of helices and sheets within a domain, but before the Foldit design, there was no suggestion that this fold could be stable by itself. 14:04
alcor29 what will happen now with the 1384 design? 14:06
smortier QUESTION: Baker lab has had some recent success in producing thousands of small protein designs computationally. Is there still a place for hand-crafted designs like foldit players produce? 14:07
bkoep @alcor, any number of things. For example, if anyone needs a stable protein scaffold for another design project (a common strategy in the Baker lab), they have an entirely new fold available to them 14:08
bkoep Absolutely! I think this question is referencing the recent Science paper: 14:09
bkoep http://science.sciencemag.org/content/357/6347/168.long 14:09
bkoep The Baker lab scientists are getting ever better at producing large numbers of design, but diversity is still a sore spot 14:10
smortier QUESTION: If you were going to add an extra amino acid to the standard set of twenty, as now seems possible, what would it be? 14:10
bkoep Those thousands of small protein designs comprise only a handful of folds. Foldit players produce much more varied structures—even with the stringent loop requirements 14:11
horowsah It is now possible to put non-standard amino acids into proteins with some regularity, but it’s still much harder than using the standard 20. So although it would be a really cool thing to eventually be able to do design in Foldit with a bunch more amino acids, being able to test the results experimentally is still too challenging on the whole. 14:13
smortier Looks like our time is up. Thank you everyone for your questions! 14:13
smortier As a reminder, chats are logged and we post the log afterwards. 14:14
bkoep Thanks all for the great questions! 14:15

Generated by irclog2html.py 2.13.1 by Marius Gedminas
- find it at mg.pov.lt!

Get Started: Download
  Windows    OSX    Linux  
Windows
(7/8/10)
OSX
(10.7 or later)
Linux
(64-bit)

Are you new to Foldit? Click here.

Are you a student? Click here.

Are you an educator? Click here.
Search
Only search fold.it
Recommend Foldit
User login
Soloists
Evolvers
Groups
Topics
Top New Users
Sitemap

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