Developer Chat

@inkycatz Okay everybody! We're just about ready to get underway with our chat today! 14:00
@inkycatz First some introductory remarks, many of which I typed in advance.... but first I want to thank everyone for coming. 14:01
@inkycatz And for the great enthusiasm in posting questions for us to look at on our thread. 14:01
@inkycatz Welcome to today's chat! We have a number of questions to go through, so if you can just be patient (including while people may be typing!) we will try to get to them as soon as possible. 14:01
@inkycatz Today's topics were chosen by all of you and include (but are not limited to) wiggle power, scripting, and puzzles. 14:01
@inkycatz Please keep your questions short and direct as possible so we can keep things moving and be courteous to everyone else, we have a lot to go through. 14:01
@inkycatz Global is a busy place, and while we love sidechatter and questions, if we can keep those to a minimum during our chat, I promise you can have it back soon. :) 14:02
@inkycatz (usually, seems pretty chill today tho) :) 14:02
@inkycatz Although we would love to have feedback on a lot of topics, let's keep this to direct questions only. The forums and feedback area are a great place to expand your thoughts in great detail and I encourage you to use them liberally for that purpose. :) 14:02
@inkycatz Please be aware although we cannot reply to all feedback and forum threads, they are being read! 14:02
@inkycatz And they are appreciated. 14:02
SethCooperIRC hi everyone 14:02
@Susume2 hi Seth 14:03
@inkycatz If you have a followup question on a question, lets try not to get derailed on a topic endlessly. We have a good number of questions today and I want to make sure we get as many of the posted questions as we can. :) 14:03
KarenCH hi! 14:03
@inkycatz Speaking of posts in the thread, not everything will go in posted order - several topics were clustered together. 14:03
@inkycatz If you do not get your question answered today, please add it to the thread when the chat log is posted. 14:03
@jflat06 hello! 14:03
@inkycatz Again let's just get this party started! We will also be having chats in October and December :) 14:03
@inkycatz yay! 14:03
@silverberg hello 14:03
@betahelix Hi everybody! We also wanted to start by thanking you all for your hard work during CASP11!!! :-) 14:04
@TimovdL hi all 14:04
@inkycatz Thanks beta! I knew I was forgetting something essential. :) 14:04
@inkycatz Let's start with some wiggle power questions. 14:04
@inkycatz first up: is high wiggle power necessary? 14:04
zlatankg what's a wiggle power? :D 14:05
@inkycatz (sorry! waiting for jflat to get back to his desk for a sec ;) 14:06
@inkycatz I will say in the  meantime we do have a lot of great info on wiggle power in the community and any newcomers unfamiliar with the lingo we're tossing around 14:06
@jflat06 We did a lot of testing on this pre-CASP - The comparative testing that we have done did not show any actual scientific benefit to allowing high power wiggle. 14:06
@inkycatz feel free to ask those questions after :D 14:06
@inkycatz Next up! 14:07
@inkycatz "On a recent puzzle, I got a ludicrous number of points (> 200) just on medium wiggle, but it took all day, with local wiggle giving < 0.001 points per iteration. The overall shape of the protein didn't seem to change at all. Is ideality doing what it's supposed to, or is it just lowering the score without reflecting changes in the structure?" 14:07
@jflat06 that said, we allowed high-power during the CASP refinement puzzles, and are continuing to evaluate its usefulness before ruling it out. 14:07
@inkycatz (oops sorry, jflat, thought you were done :) 14:07
@jflat06 For ideality, it's mainly there to keep the protein from getting too wacky. However, it does allow for some optimization as well. A protein doesn't have to be *perfectly* Ideal - it just has to justify being unideal by getting some benefit elsewhere. 14:08
@TimovdL Can there be a middle ground between high and medium? 14:08
u16985753 Are you technically complete with something whenever you decide you're done and have the score you want? Or are there specific objectives other than removing voids and clashes? 14:08
@inkycatz (you can just done at me when you're done, jflat :) 14:09
@Susume2 u16 in tutorials there is a target score; in science puzzles just try to get highest rank you can before they expire 14:09
SethCooperIRC technically there might be something between medium and high 14:09
SethCooperIRC but it's not clear it would be useful 14:09
BletchleyParkirc How good is the scoring function then if there is no scientific gain in using high wiggle yet high wiggle gives a lot of points ? 14:09
@jflat06 @Timo - high and medium are different sets of degrees of freedom that are allowed to vary. Medium includes extra DOFs for the backbone, high also includes the sidechain DOFs. You could imagine including SOME sidechain degrees of freedom as a middle ground, but it's not clear that would be useful. 14:10
@jflat06 @BP - in this context, we evaluate usefulness in terms of differentiation of good/bad structures. 14:10
u16985753 Thank you! 14:10
@jflat06 so score relative to RMSD or GDT 14:10
@jflat06 alright, next question? 14:11
@inkycatz Has something happened to auto wiggle power? Auto didn't get a good reputation at the start, but I've recently had encouraging results, including a successful solo on CASP puzzle 957, which spent a long time on auto, by mistake. 14:11
@jflat06 This is good to hear. We made an update ( to the game part way through CASP that addressed some discrepancies between how ideality was scored and how wiggle and idealize behaved. With that update, Auto-wiggle should behave much more like Low-power unless you have marked unidealities. 14:11
SethCooperIRC yeah 14:11
@jflat06 The scoring issue was causing auto to behave more like medium than it should have 14:11
@TimovdL The problem I see is that the current high almost freezes the protein, if that could be prevented, while removing bad ideality and backbone scores I would welcome that 14:11
aramn76 the hole idea of the game is that you as a computer will move the proteins and strans until a solution comes into place is not that for you to learn then the info go to the lab server and its shows the multiples attempts 14:12
@jflat06 @Timo - high power wiggle optimizes very fine degrees of freedom. This pretty much results in a frozen protein, no matter how you do it. 14:13
@jflat06 High power can actually introduce extra unideality, since it's allowing more degrees of freedom to vary from their ideal values. 14:14
@jflat06 Running medium or auto-wiggle is intended to help you resolve very-bad idealities, such as when a cut is closed. 14:14
@jflat06 if you want to completely get rid of unidealities, the idealize tool is probably the best bet 14:14
@jflat06 alright - should we move on? 14:15
@inkycatz Next up we have a smattering of questions relating to specific feedback and ideas. It's also the time where I do my community things to remind everyone that priorities and issues change over time, and a "we'd like to" is not a set in stone promise. ;) 14:15
@inkycatz So be kind when you read the chats later. 14:15
@inkycatz This one is about Do you have any feedback on the thought for having LUA commands to set and get the wiggle power? 14:15
SethCooperIRC this is something that would be nice to add 14:16
SethCooperIRC now that it seems wiggle power is a useful part of the game 14:16
alcor29 So multiple wiggle powers are here to stay? 14:16
@jflat06 @alcor - at the very least, low/medium/auto are here to stay. 14:17
@jflat06 low and medium have actually always been in the game 14:17
SethCooperIRC that's true 14:17
@jflat06 they were just hidden from players and would automatically alternate between low and medium when you ran a wiggle 14:17
alcor29 why not go back to that then? 14:18
spmm2 O.O!!! alcor 14:18
KarenCH ohhhh is that why there are claims that wiggles should be done in multiples of two iterations? 14:18
SethCooperIRC yes 14:18
@jflat06 we chose to expose them because this gives players much better control, while making the actual behavior of wiggle clearer. 14:18
@TimovdL So we need sev eral functions, one to known which ones are available, seterr and getter 14:18
@jflat06 @Karen - yes, that is the historical reason for doing wiggles of 2 iterations 14:18
alcor29 true but we have to run the same scripts at diff wps. 14:19
SethCooperIRC if a script didn't set wp, then it could run at whatever it's set to in the gui 14:19
@jflat06 it actually turns out that you can normally use low-power for most of your folding 14:19
@jflat06 which gives us much better performance for wiggle than alternating every other iteration 14:20
spmm2 :D 14:20
@TimovdL most of my folding is 14:20
@jflat06 that was another reason for switching to letting you choose which wiggle power you wanted. 14:20
@jflat06 alright - next? 14:20
alcor29 yes but in the competitive environment you have to go to med  to get more points 14:20
@inkycatz We've got a few questions relating to scripts! 14:20
@inkycatz This one's about Are there any plans to add or implement LUA commands like band.SetNote and band.GetNote? 14:21
SethCooperIRC there isn't a way to set a note on a band in the gui now is there? 14:22
@jflat06 Notes on bands could be a useful feature. Adding that in, with a proper interface isn't trivial, and so we may end up doing it when we have more time. 14:22
@inkycatz Ok, moving to the next q 14:23
@inkycatz Currently, some high scores get "stuck" with an open cut. The cuts could be easily closed by familiar scripting actions in most cases. At present, a script can't detect the open cut. Manually detecting and correcting this condition means lost hours and missed opportunities.  Will scripting be given support to handle cuts in the future? 14:23
SethCooperIRC right, we'd have to add them in general before scripting 14:23
@inkycatz sorry Seth, didn't mean to cut you off there :) 14:23
@jflat06 For a function for getting the current set of cuts - this is something I'd like to add. 14:24
@TimovdL A penalty for open cuts would prevent polution of recentbest 14:24
@jflat06 It should be in the game, and can probably be added without too much effort. 14:24
@TimovdL Also a function Canbeclosed would be useful 14:25
jeff101 I think RMSD and GDT focus on backbone coordinates. Perhaps the small change in them while the score rises is due to sidechain changes. Do you ever use a measure that includes sidechain positions? 14:25
@jflat06 @Timo - one suggestion was just to have the close function return whether the close was successful, which could serve that purpose as well. 14:25
spmm2 penalty can't to be too large as we have to monitor score working with open cut for for hours 14:25
KarenCH it wouldn't completely solve problem - because cuts may have been there already 14:26
spmm2 band open cuts as one command would be useful 14:26
@TimovdL That would help also, but still make recentbest poluted 14:26
@jflat06 @jeff - RMSD and GDT are general comparisons of a list of points, and can be applied to the full-atom pose as well. 14:26
@jflat06 or at the very least, RMSD can. 14:27
@jflat06 i believe GDT is more specific. 14:27
@betahelix @jeff101 we used all-atom RMSD for the core residues in the first Nature paper, but generally just look at the backbone 14:27
@TimovdL Or there should be a pose recentvalidbest 14:27
KarenCH (I still want close to return whether successful - just also I want to be able to ask "is there a cut at X"? 14:27
KarenCH there is a creditbest pose 14:28
@jflat06 @Karen, we are interested in adding both of those. 14:28
KarenCH :) 14:28
@jflat06 next question? 14:28
@inkycatz Regarding - What are your thoughts on letting players designate models/tracks that are still improving to be automatically shared with scientists when a puzzle ends? 14:28
SethCooperIRC that could be useful 14:29
jeff101 would be helpful in multistart puzzles. you do want diversity of solutions, don't you? 14:29
bkoep I posted a reply to that feedback 14:29
@jflat06 This could be an interesting idea, but would require a lot more investigation on how to set up a system for doing so, and making that system useful. 14:29
SethCooperIRC how to determine if it is still improving? 14:29
bkoep I'm not convinced we want to automate anything to Share with Scientists 14:29
jeff101 right now best evo and best solo get automatically shared when a puzzle ends 14:30
@jflat06 @bkoep - i agree that 'share with scientists' should be manual exclusive. however, you could imagine a different system for this. 14:30
jeff101 why not let specific tracks get shared too? 14:30
BletchleyParkirc What if they improve after puzzle end ? 14:30
@TimovdL One way to help to circumvent this is to keep the server open for shares for a day after closing 14:30
jeff101 automating would have helped a lot during CASP 14:31
jeff101 many times we had to e-mail to make sure our shares were included 14:31
@betahelix @jeff101 at least we have 2 years to get an easier system ready ;-) 14:31
@jflat06 the flip side of some of this is that the end-stages of folding are usually optimizations of existing folds 14:31
@jflat06 as opposed to structural changes 14:31
@jflat06 so they will not help us sample the space any better. they could potentially increase our ability to differentiate good/bad solutions, though. 14:32
jeff101 so one alternative for players is to share with scientists more often ... does that really help? 14:32
drumpeterIRC hey guys! i made it! :D 14:32
@inkycatz (welcome!) 14:32
@jflat06 @jeff - we have found share with scientists to be most useful in design (and of course debugging) 14:32
@Susume2 during casp they matter to the teams that are submitting their own soluions, as they allow alternate solutions to be returned to the team 14:33
drumpeterIRC agreed 14:33
@jflat06 in design, the score function is less able to properly compare one structure to another than in prediction 14:33
vmulligan I don't quite understand -- why would it be useful to automate sharing with scientists? 14:33
@jflat06 i'm not sure it's necessarily 'sharing with scientists' as much as it is allowing players to continue working on puzzles post-expiration. 14:34
@betahelix @Susume of course! which was a feature we were excited to have implemented in CASP this time around. @vmulligan it would have made it easier for CASP11... 14:35
@TimovdL Because sometimes there are changes (good ones) that work out better and people are asleep or at work and cant share it after closing 14:35
jeff101 best evo and best solo automatically get shared with scientists. what if a player is runnng multiple distinct tracks based on multiple starting positions for example. It would be good to have each of those lesser-scoring tracks get shared with scientists too. 14:35
@jflat06 ah - i see your point. 14:35
@jflat06 you are saying that you have different approaches to the same puzzle, and you want to highlight the end result of each of those approaches. 14:35
@jflat06 that could be interesting, yes. 14:36
drumpeterIRC is this scichat about the results of CASP and what's new? 14:36
@inkycatz This is a devchat but we'll talk casp results in future chats I'm sure :) 14:36
@TimovdL On the current design one I have 2 different ones, scoring about equal 14:36
@jflat06 they are actually all shared with us, and are included in our global analysis of the puzzle 14:37
jeff101 puzzles with mutliple cysteines ... I like to try models with each of the disulfide bridges that are possible 14:37
@jflat06 but we could do a better job of highlighting players top attempts at different approaches 14:37
@jflat06 alright - we should probably move on. 14:37
spmm2 any chance of fixing the bug voids causing crash if showing when client minimised 14:37
drumpeterIRC Timo, where you able to get more than one helix? 14:38
@inkycatz We're going to hit two tech questions and then move on to some puzzle questions. ;) 14:38
@inkycatz The deaf client issue is still open. Is this going to be addressed in any fashion? 14:38
@TimovdL No differnt nr of sheets 14:38
@jflat06 I'm not totally sure what is meant by 'deaf' client. 14:38
BletchleyParkirc Client that is not connected to server and will crash when trying to reconnect 14:39
BletchleyParkirc indicator light says 'green' 14:39
@TimovdL Not getting updates while showing it is still connected 14:39
@jflat06 Our connection between clients and the foldit servers is not stateful. Any time the clients wants to send something to the server, it connects at the point of the request. 14:39
@jflat06 so in that sense, the client cant be "disconnected" 14:39
BletchleyParkirc unless the connection was interrupted and it thinks it is connected while it is not. 14:40
@jflat06 however, it may fail to send, and should update the indicator light when it is unable 14:40
BletchleyParkirc It does not 14:40
@Susume2 if it is not getting periodic updates from the server about other people's scores, it is efectively disconnected 14:40
@TimovdL How come when the client is like that it hangs when the connection is required? 14:40
@jflat06 but we have not been able to reproduce that 14:40
BletchleyParkirc It happens every week here 14:40
BletchleyParkirc to my great irritation 14:40
sloIRC Quick note on topic of casp. The casp folks have been holding off on many of the natives. Today is the last day of casp, we are expecting some of natives to be released (After 5pm today (pacific time)). **excited** 14:40
@TimovdL Mostly on long running ones 14:41
BletchleyParkirc This is te issue: 14:41
SethCooperIRC it might be waiting for a response from the server when it is trying to connect 14:41
@jflat06 @Susume - you can think of it that way. 14:41
jeff101 sometimes many shared solutions on my team have usernames --- 14:42
BletchleyParkirc it will not detect that, it can run for days without noticing that 14:42
SethCooperIRC you can try to enable timeouts in the login dialog 14:42
@jflat06 it's not so much disconnecting as repeatedly failing to connect to the server 14:42
BletchleyParkirc and nothing gets sent to the server 14:42
drumpeterIRC i must say casp was pretty fun this year 14:42
@inkycatz Seth, jflat, what data could be helpful to pinpoint this issue for investigation later? :) I don't want to get too hung up on this one tech issue :) 14:42
BletchleyParkirc I have timeouts disabled since 2012 14:42
@TimovdL And nothing gets send from the server 14:42
BletchleyParkirc This is a very significant issue, as no science data gets sent to the server when it happens. 14:43
@jflat06 @inky a reproducable context where we can get the client to repeatedly fail to connect to the server 14:43
@inkycatz Okay. 14:43
BletchleyParkirc unplug the metwork cable for a few minutes, ten reconnect ? 14:44
@jflat06 we have tried that 14:44
@jflat06 along with a number of other connection loss situations 14:44
@TimovdL Would it help to get this error sorted to have a life client form someone that has this problem? 14:44
BletchleyParkirc maybe try that via a router and leave the pc connected to the router, then disconnect the router outgoing from the network ? 14:45
@inkycatz Sounds like something we'll just have to keep discussing on that thread. :) 14:45
BletchleyParkirc and resolve. 14:45
BletchleyParkirc we can move on 14:45
@inkycatz Hopefully we can keep at this and get a solution. :) 14:45
@inkycatz Lets get a couple puzzle questions in. :) 14:46
@inkycatz There are a few CASP natives out. Did Newchapter do a better job at picking solutions than two years ago? Can we see some of the scatter plots? 14:46
@TimovdL And a mechanism to get debug info from that client in that situation (NOT in log.txt) 14:46
@jflat06 We don't have enough comparative analysis to other groups in order to draw a conclusion yet. 14:46
spmm2 how long does it take to make a scatter plot? 14:47
@jflat06 It's not just about how well we do - we have to compare it to other groups. 14:47
@Susume2 wondering about the score vs RMSD and score vs GDT plots; not about other groups 14:47
jeff101 but they could have improved to 14:47
@jflat06 A puzzle might be easy or hard. Comparative analysis is where we need to draw our conclusions from. 14:47
jeff101 I agree with susume. I would like to see that. 14:48
SethCooperIRC i am also curious how contacts helped 14:48
jeff101 It would help assess how well Foldit and our teams choses models to submit. 14:48
vmulligan It's a bit risky to draw conclusions before all of the data are in.  Jflat06 is right -- saying, "Here's our result" isn't meaningful without being able to say, "Here's our result in comparison to other approaches". 14:49
spmm2 contacts were great - but there can be too many 14:49
@betahelix The good news is that the CASP organizers do a very thorough analysis of all the CASP targets, and post the results on the CASP website, but it does take them a very long time. 14:49
alcor29 Can you compare accuracy as ooposed to the previous foldit casp10? 14:49
sloIRC the problem of comparing newchapter to "oldchapter" using casp puzzles, is that the targets are very different between casp10 and casp11. But we could rescore the solutions (from both casps) using the different energy functions and see if the plots differ. 14:49
@TimovdL And even at 25 the reward it to high for a contact 14:49
@jflat06 @alcor - you could try, however those are a different set of proteins, and one set might be easier than the next 14:50
jeff101 @Timo. What do you mean? 14:50
spmm2 score bonus for contacts is too high 14:50
@jflat06 it's better to wait for a bigger picture before drawing conclusions 14:51
@inkycatz We ready for a new question? :) 14:51
@TimovdL 25 pts for a contact made makes the protein not really a nice one 14:51
@jflat06 yeah 14:51
@inkycatz Another Susume question! Then we're going to go with a science question and I'm going to open it up to our devs and scientists to ask any questions they have for you guys. :) 14:51
@inkycatz When will ED puzzles be back? 14:51
jeff101 many models were chosen based on NewChapter scores. Did high NewChapter scores give better RMSD vs. Native than CASP10 had. 14:51
@jflat06 We need to resolve some scoring and behavior issues before we can post new ED puzzles. 14:52
@inkycatz Okay 14:52
@jflat06 Some of the changes in newchapter introduced some problems with ED. We need to re-evaluate whether those issues still occur and get fixes for them if they do. 14:52
@TimovdL Feel free to do some experimenting in devprev with that, I like ED 14:52
@jflat06 we can probably get some of that started 14:53
jmbrownlee333     I would love to know better what a contact means. Does it mean something different for True contact puzzles. Are they fuzzy things or determined from data, like NOE's? 14:53
jeff101 will there be any symmetry puzzles soon? is symmetry still sort of broken? 14:53
spmm2 why are freeform design puzzles so rigid? 14:54
@inkycatz (these are all great questions, folks) 14:54
@jflat06 @jeff - there are still some issues with symmetry. they're very hard to track down, but we're eager to start running symmetric design puzzles again. 14:54
SethCooperIRC yeah, symmetry and density are ones we are interested in getting running 14:55
@jflat06 @jmbrown - sometimes they are predicted by automated methods, sometimes they actually do come from experimental data 14:55
drumpeterIRC yea seth, I miss symmetry 14:56
@Susume2 the casp ones were from data 14:56
@jflat06 should we finish up the last scheduled question and then go to open questions? 14:56
spmm2 can we have all the categories showing on our player pages? Spent 109 days doing 73 puzzles and CASP11 is not showing 14:56
@inkycatz Yep! 14:56
@inkycatz Ordinary wiggle and banding actions generally manage to cause sheets to appear "naturally" (as far as our processes are natural), but helices don't do so - if there's a stretch of aas that ought to become a helix, no amount of ordinary actions seem to cause helixing to happen spontaneously. But in real life proteins, helices really do exist. What's going on, what's the reason? What kinds of 14:56
@inkycatz actions would encourage "helixing up" to occur? 14:56
@jflat06 In-game actions and the Foldit environment are not a physics simulation. The behavior of the Foldit environment may encourage some types of secondary structure formation, but physics behave differently. In Foldit, we give you tools to specifically accomplish things like forming helices. 14:56
@inkycatz After this, I'd like to open the floor first to the devs and scientists to ask all of YOU questions. :) 14:56
vmulligan @spmm2:  By "rigid", do you mean, why do we give you so many constraints related to secondary structure and so forth?  Part of the reason for that is that Foldit/Rosetta's energy function is calibrated using naturally-occurring proteins, with lots of secondary structure.  We're most likely to have pathological scoring behaviour far from structures like that -- i.e. there might be ways to get good scores with things that are very loopy, but w 14:56
@inkycatz after all, we've been asking them a lot, we should give them a turn 14:57
@inkycatz then we can just open up the floor generally :D 14:57
@jflat06 i guess my biggest question would be about tool behavior. have most of the behavior issues been resolved since newchapter came out? 14:57
sloIRC @jmbrownlee333 the sparse contacts are from simulated NMR data (NOE). The true contacts are a subset of predicted contacts that are verified to be correct. For the True contact puzzles, the non-ambiguous contact from the NMR data is used along with the correct ones.  The sparse contacts can be "fuzzy" in that the distance could be > 8 between CA-CA. 14:58
@jflat06 tool behavior is a very fuzzy thing, and hard to get concrete test cases for. 14:58
jmbrownlee333 thx 14:58
spmm2 low wiggle power has fixed all the problems I used to have with changes being made to wiggle 14:59
jamiexq I got a tool error just now with idealize ss on the ebola puzzle 14:59
spmm2 ebola is closed? 15:00
spmm2 I shut mine down 15:00
jamiexq 971 is closed? 15:00
@jflat06 are there any other last minute questions? 15:00
KarenCH brand-new just opened one. not the original version of the puzzle 15:00
@inkycatz 971 is open 15:00
@betahelix we can change it to 971b, but don't want to confuse people even more! 15:00
@inkycatz hah :) 15:01
vmulligan 971 is back.  That was my mistake -- apologies.  The inital version had a problem with it. 15:01
@inkycatz vmulligan, beta, bkoep? any questions you all have for folks? 15:01
drumpeterIRC may i bring something up? 15:02
Alreadytaken As a new player I find that there is a big gap in difficulty between the tutorials and the puzzles will there be more easy puzzles? 15:02
spmm2 calling it a b is standard practice for a reopened puzzle 15:02
@jflat06 @Alreadytaken - yes! we normally post 'revisited' puzzles, which are easier. However, with the CASP competition going on, we didn't have much room for these puzzles. 15:02
spmm2 and it is obvious in the notifications 15:02
jeff101 I want to rephrase an earlier question: did the highest NewChapter scores give the best RMSD and GDT scores in CASP11? 15:03
@inkycatz (If not, we can close the official part of chat and just move on to the informal stuff. FYI, appreciate you all sitting through a lot of questions, but mostly I appreciate the posting of them in advance!) 15:03
@betahelix now that CASP and its mighty puzzle load is over (thanks again for everyone's patience during those crazy weeks!) how many Science Puzzles up a time is ideal for you? 15:03
drumpeterIRC i feel that we haven been getting to many puzzles at one time 15:03
jamiexq at minimum 3 (non-beginner) 15:04
@inkycatz (Also I appreciate you all for coming - look for chats coming up in October and December!) 15:04

Generated by 2.13.1 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.
Social Media

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