Developer Chat

bkoep Hello all, and welcome to our Science Chat! 14:01
bkoep It's been a while since our last chat, and we're happy to answer any science-related Foldit questions from players 14:01
bkoep Since we had such short notice for this chat, and we didn't receive a lot pre-submitted questions, this chat will be more "open" than previous chats 14:03
spvincent I submitted a rather general query about a new development in CryoEM 14:03
@betahelix Yes, great question! 14:04
@betahelix It's a very exciting field, as that area is rapidly developing. 14:04
@betahelix For now crystallography will probably still dominate small proteins for several years... 14:05
@betahelix but in the more long term... EM might take over! 14:05
@betahelix We might be using cryo-EM to solve your Foldit designs in the future :-) 14:06
bkoep By the way, spvincent's original question is here: 14:07
bkoep https://fold.it/portal/node/2006106#comment-37701 14:07
bkoep It concerns some recent exciting reports about using a special cryo-EM technique to solve the crystal of small molecules (much smaller than proteins) 14:07
spvincent It might be nice to have some small EM-derived structures to work on. The current big protein is a bit intractable for me. Not to mention being hard on the eyes. 14:07
bkoep *crystal structures of small molecules 14:07
@Susume2 the first EM puzzle we had was several years ago and the resolution was terrible, but the resolution is great on the current one 14:08
@betahelix That was going to be my question to everyone: how are you all handling the 221-residue ED puzzle! 14:09
spvincent I would like to like them, but its just too much strain on the eyes 14:09
@betahelix No just in terms of the resolution... but in terms of performance. 14:09
@betahelix *Not 14:09
jflat06 I will note that we would really love to give you smaller ones (and we may for intruction purposes), but as far as true blind cases go, we have to take what we can get! 14:10
rmorettiIRC (Anything we could do to help with eyestrain issues?) 14:10
Wbertro performance is great here, except for the crashes with not line info in log.txt 14:10
spvincent performance not too bad 14:10
jflat06 That's actually surprising to hear. 14:10
@betahelix spvincent, do you mean eye strain due to the density.. or just a large protein chain 14:10
@Susume2 performance is good for me too, but I'm on a fairly new gaming machine 14:11
@betahelix but we are sooo happy to hear it! 14:11
jflat06 Right, I'd imagine it may be straining some older machines. 14:11
Wbertro on devprev 14:11
bkoep We also have an upcoming release (now in devprev) that we hope will speed up symmetry puzzles 14:12
Wbertro nice 14:12
spvincent I just find them hard to look at. You provide plenty of ED dispaly options so maybe I'll keep trying till i find some combination that works for me 14:12
jflat06 We do have a lot of options, but there's always room for improvement on that front. Do people find the existing options helpful? 14:12
jflat06 I have read some of the suggestions for ED, and they are good (we have thought about some of them internally, even) 14:13
@Susume2 shift-Q and the mouse methods for trimming foreground and background are most helpful to me - oh and the pins! 14:15
jflat06 Good to hear the pins are working well with that! 14:16
jflat06 By mouse methods, which feature are you referring to? 14:16
@Susume2 ctrl-shift-drag up/down and ctrl-alt-drag up/down 14:16
@Susume2 they fade foreground and background respectively (not just on ED, on anything) 14:17
@Susume2 oops I switched them 14:17
jflat06 Hah! I hadn't realized someone had implemented that. That may be a feature worthy of the ED tutorial, especially if you are finding it useful. 14:17
@Susume2 oops, its shift-alt-drag for foreground 14:18
jflat06 Would it be ideal to be able to do this for *only* the density, instead of density+protein? 14:18
@Susume2 if you can add that without taking the other away, it would be great 14:20
jflat06 yeah, I don't think we'd want to remove the existing feature at all. 14:21
@Susume2 maybe sliders on the ED dialog (send RGB sliders to a "more" dialog, those hardly ever need changing) 14:21
Wbertro agreed 14:22
jflat06 That's a fair point, is the RGB even useful? 14:22
@betahelix But leave Alpha and Threshold? 14:22
@Susume2 yes alpha and threshold both useful 14:23
@betahelix great! 14:23
Wbertro I work with RGB = 0,1,1  so yes 14:23
@Susume2 people do have their favorite colors, they would be sad if you took them away entirely 14:24
Wbertro but other dialog = very good 14:24
jflat06 Ok, cool. Yeah, I think moving it into a collapsable menu or something like that would be ideal. 14:24
jflat06 It does take up way too much room for how often it is accessed. 14:24
Wbertro also None, Solid and Wireframe = ok, others for me not so 14:25
@Susume2 same: none, solid, wireframe are enough for me 14:26
jflat06 It would be cool to have volumetric rendering in Foldit as well, but I have a feeling that would kill our performance (if any of you have played Mozak, that's volumetric rendering) 14:26
@betahelix Do you feel that if we improved how you can view the ED, you'd be able to play larger ED puzzles? (say ~250 residues) Or is it all processing slowdowns at that point. 14:27
@Susume2 can't play mozac, performance too bad ;) 14:28
jflat06 haha 14:28
jflat06 Yeah, it's a bit demanding 14:28
@betahelix (I'm not saying we're even considering posting such puzzles, I'm just curious) 14:28
Wbertro maybe it would be great to have a switch to attache the shift-q behavior to the left-mouse click on a segment to follow you around... 14:28
Wbertro *attach 14:29
jflat06 Basically, to let you move the protein/density together? 14:30
Wbertro when exploring the wireframe / segment interaction 14:30
@Susume2 shift-Q is center, zoom, and cut-away all in one 14:30
Wbertro I have to get close to the segment with shift-q 14:31
Wbertro maybe an alternate of shift-q where you get close to the segment and can view its relation to the wireframe 14:31
@Susume2 I can't play ED without it 14:32
Wbertro the problem is when you start to move away from the current shift-q segment, you start loosing the wireframe 14:32
Wbertro and need to redo the shift-q on that next segment 14:33
rmorettiIRC Are you doing this in Selection interface or in the Original interface? 14:34
Wbertro but the centering I could do without, just keep me basically where I am with the density / protein 14:34
Wbertro I use Selection interface exclusively 14:34
@Susume2 selection 14:34
jflat06 So the trim you're referring to here is the trim foreground/background you mentioned earlier? 14:36
Wbertro also using foldit on 2 screen where the second screen width is only used 2/3 of.  So my center is offset . 14:36
jflat06 Do you find the "trim density" option useful? 14:36
Wbertro not much for me 14:37
Wbertro cannot adjust it, it won't remember better clicks so cannot adjust it 14:38
Wbertro slider is always at Far when it opens 14:38
@Susume2 yeah slider doesn't remember your setting, so you can't adjust from your current setting 14:39
Wbertro but even if I set it to near, it makes almost no change on my PC 14:39
bkoep Can I ask what everyone thought of the recent Protein Design Partition Tournament? 14:40
bkoep Was it enjoyable? Was it informative? 14:40
Wbertro I liked it and very informative for me 14:41
robgee me too 14:41
@Susume2 very cool dynamic with defense round and offense rounds, weekly feedback made it informative - without the feedback it would be much less fun 14:41
bkoep For those of you that submitted designs in phase one of the tournament, did the tournament affect your design strategy at all? 14:42
robgee yeah, the weekly feedback was good 14:42
bkoep Or did you essentially treat phase one just like a regular design puzzle? 14:42
spvincent I pretty much treated it as a regular design puzzle, although looking back I should have done things a bit differently 14:44
@Susume2 tournament suggests I should pay more attention to score and less to cool/creative (probably not what you wanted to hear) 14:44
Wbertro I did not participate in first round due to real life, but it will definitively affect my regular designs 14:45
jflat06 Do you still feel it will affect your designs, even if you know other players wouldn't have a chance to attack them/ 14:45
jflat06 *? 14:46
Wbertro and also strategy for finding alternate folding solutions 14:46
Wbertro yes 14:46
@betahelix I was also curious as to how everyone held up with the puzzle load. Even during CASP we never had that many puzzles up at a time. Did it work out for you? Were the blog updates enough for you to decide on which one to play/focus on? 14:47
spvincent It was a bit too many for me 14:48
jflat06 Do you feel like it was still worth it, in terms of what you got out of the experience? 14:49
Wbertro I needed more time to find alternate folding position and evaluate them 14:49
jflat06 Would it be worth ever doing somethign like this again? 14:49
Wbertro for me, yes 14:49
robgee yes, but not too often :) 14:50
@betahelix So once a month?;-) j/k 14:51
jflat06 haha 14:51
spvincent How much extra work is it on your end? 14:51
jflat06 Yeah, I can definitely appreciate that it was a bit much. 14:52
bkoep That's helpful feedback! The tournament definitely took up some of our time, but I think it was a worthwhile experiment. 14:53
bkoep We've been trying to think about how to improve information flow in Foldit—especially between players. 14:53
bkoep For example, I've always been impressed by the sharing of screenshots after a puzzle closes. 14:53
jflat06 One thing we may consider in the future is a system to allow these types of puzzles in a cleaned-up format, which may make them significantly easier to deal with, instead of having a ton of individual puzzles. 14:54
Wbertro I would like to be able to get the recipes to have a LUA function returning the current delta from the target 14:57
Wbertro the Exploration "score" 14:57
bkoep @betro, yes! We're sorry there wasn't a good way to get the RMSD value in LUA, during the tournament. 14:58
spvincent Maybe you should start with an extended chain and have no knowledge of the submitted structure 14:59
Wbertro no problem, if you can do it sometime... 14:59
Wbertro extended chain requires more time also 15:00
jflat06 That's something we'll keep in mind if we end up doing the tournament again. Thanks! 15:00
bkoep @spvincent, that's a good idea! 15:00
jflat06 Yeah, I'd imagine extended chains would be brutal. 15:00
betahelix That's a really cool idea, though... that's what Rosetta@home does! 15:00
bkoep Do you think you would have been able to find the intended designs? 15:00
bkoep ...starting from an extended chain? 15:00
spvincent it would take a lot longer but you wouldn't be biased by knowledge of the structure 15:01
spvincent However, it would be a lot more work 15:01
Wbertro mostly not always, they would definitively be different 15:01
jflat06 Do you think it would be frustrating to be working from an extended chain, but still be penalized for going near the original player's design without being told what it is? 15:01
@Susume2 the successful challenges to the 2nd fiendish ghoul protein in the blog were from extended chain 15:02
spvincent No, you'd still have have the RMSD value there 15:02
Wbertro yes, needed so we don't get there by mistake and waste time 15:03
@Susume2 getting the RMSD penalty would be an aha! I found it! moment 15:03
jflat06 Right, but I can imagine it might be frustrating to be told to stay away from a something without knowing what it is 15:03
jflat06 haha, that's one way of looking at it 15:03
Wbertro lol 15:03
Wbertro but no 20 extended chains at the same time... 15:05
jflat06 Definitely not 15:05
bkoep Well, it's back to work for me. Thanks everyone for hanging out with us this last hour! 15:08
betahelix It would be like finding an oasis in a desert, only to find out it was a mirage! 15:08
bkoep I'm sorry we didn't have as much time to solicit questions in advance! 15:08
spvincent quick q before you go 15:09
betahelix Of course! 15:09
spvincent aflatoxin 15:09
spvincent We're now on the 11th iteration 15:09
spvincent Any indication that we're homing in on something useful? 15:10
bkoep @spvincent, good question! 15:10
spvincent Although I realize you may not be able to be too specific 15:10
bkoep I've heard that the Siegel Lab is finally starting to get some experimental data about designs from some of the earlier rounds 15:11
bkoep Hopefully we'll get a blogpost from them soon! 15:11
spvincent ok, good 15:12
Wbertro by the way, we would never be able to find a solution comparable to 1588 without the ED (or maybe some contacts, but much less likely) 15:12
Wbertro maybe contacts coupled with ED would help some players find the match between ED and the chain? 15:13
betahelix Yeah, this was a very difficult target! 15:13
betahelix I'd be very surprised if any human groups in CASP were able to get near it. 15:13
Wbertro none of our design would get near that and the score would deny it from the top 100 15:14
Wbertro but it is real life stuff...;) 15:15
betahelix Well, it's particularly tricky because it's part of a much larger protein complex. 15:16
bkoep Bye all! Thanks for the questions, and the helpful feedback about the tournament! 15:16
@Susume2 do you have any sense of how many different players got above .8 GDT on the first part? 15:16
betahelix So they are all sandwich against one another (which explains how it ends up in that shape) 15:16
betahelix I don't have that data on my right now (but I'm not sure I would reveal it if I did ;-) 15:16
jflat06 Yeah, the score function without ED isn't telling the full story on this puzzle, since it's really supposed to be part of a bigger protein. 15:16
jflat06 Granted, making a puzzle of the full complex would be quite challenging, too! 15:17
betahelix We would love to give you the proteins on either side, but they are each over 300 residues! 15:17
@Susume2 no need to tell us, but some sense of how many players are getting close on the current one would tell you if current tools are sufficient for us to do large proteins 15:17
Wbertro why don't we get an ED score/bonus shown as before in 1588 15:18
betahelix I think when the full complex is solved, we should make a Galley Mode puzzle at least with the entire thing. You might not be able to wiggle anything, but (if it loads) you could see it! 15:18
@jeff101 frood was surprised the 221-residue puzzle worked on his machine, and he was able to get a really good ED score on it pretty fast 15:18
betahelix You did get the bonus.... but no bonus of interacting with the other proteins. 15:18
@jeff101 maybe a 300-AA puzzle is do-able 15:18
betahelix That's the bonus that nature has that you don't! 15:18
@jeff101 personally, one of my machines has to do less clients than usual for the 221-AA puzzle 15:19
@jeff101 my other machines seem ok with it 15:19
Wbertro we have a density score that we can calculate with recipes but no more one shown in Objectives 15:19
@jeff101 and can do the usual # of clients 15:19
betahelix Susume... I guess if it's 1 player or 100 players that made it, then it would mean the same thing... 15:19
@Susume2 yes and no, for a long time foldit did not have a very deep bench for ED puzzles, hopefully that is changing 15:20
betahelix so I can say that the score function is working exactly as we would want it to, but that's all I can say :-) 15:20
betahelix I agree with you entirely... I just meant in terms of the scoring function picking it up (just like the Partition Tournament, we don't want lower energies that don't fit the density) 15:21
betahelix (I hope that makes sense) 15:21
Wbertro got to go, wife calls, thanks all for the very informative chat 15:22
jflat06 Alright, I think we'll end the official chat here. 15:22
@Susume2 I'm not questioning the scoring function, I'm asking about the community's capacity to take on larger puzzles 15:22
jflat06 Thanks for coming everyone! 15:22
@Susume2 thanks for all your time! 15:22

Generated by irclog2html.py 2.17.0 by Marius Gedminas
- find it at https://mg.pov.lt/irclog2html/!

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