Developer Chat

inkycatz Hi everybody! Welcome to our off-schedule Science Chat! For those wondering, we’re still planning on having one in May. As you all probably are aware, chats are logged, so if you have to go don’t worry we post the log afterwards. :) 14:00
inkycatz I’d like to welcome our scientist for today, horowsah! Although we will likely be joined by beta_helix and bkoep, at least I hope so at some point 14:00
inkycatz Why don’t you kick us off in the meantime though, horowsah :D Tell us who you are, what you’re working on 14:01
horowsah Sure- some of you may remember me from the competition with the University of Michigan students 14:01
horowsah I was also involved in that puzzle, and since the foldit players did so well with that, I thought it was time to try something even harder 14:02
horowsah since you guys seemed up for a greater challenge 14:02
horowsah So we put together the puzzle that was up recently that unfortunately seemed to cause a lot of crashing issues 14:02
horowsah but should be a lot of fun, when we get all the kinks worked out 14:02
horowsah this is the puzzle where we have two binders, one of which is large and mostly fixed 14:03
horowsah and the other is mobile, and then there's some electron density 14:03
horowsah anyways, hopefully for those of you who were able to get it running it proved interesting, and we're going to tweak and repost it sometime soon 14:03
horowsah the thing that makes it a real challenge is the electron density is not so wonderful 14:04
horowsah and so finding the correct binding pose is hard 14:05
horowsah which is hopefully where you all come in! 14:05
@TimovdL Is there only one binding pose or might there be several? 14:05
horowsah there are likely several 14:06
BletchleyPark What exactly does this puzzle represent ? 14:06
horowsah so it's a crystal structure of a chaperone and a protein that it binds to and helps fold 14:06
horowsah and we know that the protein that binds it has to be able to sample multiple states 14:07
horowsah which is why the electron density doesn't look so good 14:07
horowsah it gets averaged out by the multiple different binding poses 14:07
horowsah so from my perspective, I'm hoping that players can come up with not just one good solution, but possibly several 14:08
@TimovdL So what you are really interested in is a few binding poses so you could reconstruct the sequence of it? 14:08
horowsah right, what i'm hoping for is that we'll get lots of possible solutions, and in the end it will be some combination of them that is actually the best fit to the data 14:09
horowsah as you might guess, this isn't something crystallography is typically set up to do 14:09
horowsah as usually it only is good for solving the structures of static things 14:09
jmbrownlee333 But it is ordered enough to get co-crystals to form? 14:10
@TimovdL Might it be better to restrict the possible places for us to work on? 14:10
horowsah so first on why it managed to crystallize: interestingly we only get this crystal form when we have the flexible binder present. I think it actually helps offset the entropy cost of forming the crystal. 14:11
horowsah as for restricting to possible places to work on, it 14:11
horowsah it's something that I will consider 14:11
@TimovdL Because probably the best scoring one will be the endstate 14:11
horowsah Our thought was to try and see what people get with it, and then we may restrict it down further 14:12
horowsah we haven't looked at what was done so far just yet 14:12
Skippysk8sirc many of us can't open it due to the size of the fixed crystal 14:12
horowsah yeah, we think it may have been a problem with the formatting of the map 14:12
horowsah that it got bloated when we loaded it in 14:13
horowsah that's something we're going to look into to try and figure out how to prevent that from happening before we re-post it 14:13
BletchleyPark not due to the size alone, the problem occurs especially when it tries to load the  'best ' scores saved 14:13
horowsah hmm, I will have to talk to bkoep about that- he may have some ideas 14:14
BletchleyPark you also could not load saved solutions 14:14
bkoep @BletchleyPark, yes, we did find a bug that was causing trouble with loading solutions 14:15
BletchleyPark ah ! :) 14:15
brunokestemont this is fixed in new devprev 14:15
BletchleyPark is it fixed in the client that just went live ? 14:15
BletchleyPark thank you 14:15
bkoep Yes, the bug was relatively easy to find and fix (thanks to some very helpful feedback from players) 14:16
@TimovdL Apparently not for old saves bruno has an example of that 14:16
BletchleyPark that would be a pity 14:16
BletchleyPark but back to topic. 14:17
brunokestemont I can load old saves, but the score is infinite negative 14:17
brunokestemont so with all our "best" solutions, you'll take a kind of statistic best fit ? 14:18
horowsah that's the idea- not entirely sure yet how that will work 14:19
horowsah we'll probably have to try a few different ways of mixing and matching them 14:19
brunokestemont but how would you select rank1, rank2 and other science shares? 14:20
horowsah i think we might look for structures that are as different as possible yet still have good foldit scores 14:21
brunokestemont Other question: I've the impression that what we add is outside the density map 14:21
horowsah so you should be able to turn the contour level way down- again the density for the flexible binder is quite weak 14:22
brunokestemont actually, I didnn't use the density map unless for "freezing" what was already inside of it 14:23
BletchleyPark I did not use the density map either 14:23
horowsah that's one way that could conceivably work- some parts of the flexible binder will have decent density, and other parts might not have much at all that correspond well 14:24
Skippysk8sirc would it be best to focus on stable fold that binds tightly rather than ED? obviously ED won't work for part that varies a lot 14:24
jmbrownlee333 Would different maps be useful to us? Like the fo-fc map. Or would that be too confusing, or drive the software even crazier. 14:24
@TimovdL Is it possible to down the density score of the fixed parts and up anything that is outside? 14:24
horowsah first on the question of stable fold vs ED: my bias as an experimentalist is to focus more on the ED and to sacrifice stable folds. However, I can't say for sure how it will work best and what foldit prioritizes 14:25
horowsah as to fo-fc map: it's really confusing looking. I have a hard time making heads or tails of it myself, so I think that it would likely not be all that helpful 14:26
@TimovdL So there is more delta in the density scores outside 14:26
horowsah and on the density score- we can look into tweaking it 14:26
horowsah i'm not sure how hard that would be to make it so that we up the score on things outside the fixed parts 14:27
jeff101 what is an fo-fc map? 14:28
horowsah ah, an fo-fc map 14:28
horowsah it basically shows you what still needs work 14:28
horowsah so it will show you places of positive density that need more atoms put in that region 14:28
horowsah and areas of negative density where you have too many atoms 14:28
horowsah and the idea is at the end, you shouldn't see anything at all in that map if your model is built correctly 14:28
horowsah so in this case, it in theory should show positive density, and it does, but it's really noisy 14:29
horowsah because of the multiple conformations of the binder 14:29
betahelix and implementing it would slow Foldit even more :-( 14:29
brunokestemont ok ED is noisy enough for me 14:29
@TimovdL Too many atoms or the wrong kind of atoms? 14:30
brunokestemont I used PSIPRED to predict the structure of the mobile part, it gave me helix-loop-helix just like the fixed part. Why didnt' you give us a prediction? 14:31
horowsah it could be either. If the atoms you put in are too big (like say its a sulfur modeled in where a carbon should be) it would look somewhat similar as if you had modeled in a carbon where there's nothing at all 14:31
jmbrownlee333 Its all about the electrons. 14:32
horowsah yep, it's all about the electrons 14:32
Skippysk8sirc we all know foldit can have it's scoring quirks.... how did our best folds do on the last ED puzzle? 14:33
bkoep @bruno, we didn't have a strong reason for leaving out the PSIPRED predictions—we may very well provide them in the next puzzle 14:34
horowsah to Skippysk8sirc's question, if you mean the competition with the UofM students, the best folds did really really well 14:35
horowsah we also gave the puzzle to some professional crystallographers, who did almost as well as the best foldit solutions 14:35
Skippysk8sirc so perhaps our scoring model will work well enough here if we can get this puzzle working :) 14:36
horowsah that's the hope- the competition with the UofM students makes me think that foldit might be really powerful for crystallography puzzles both easy and difficult 14:36
jmbrownlee333 esp if Susume is playing :) 14:38
jeff101 spread the word ... it would be good to have more real-world puzzles to solve 14:39
jeff101 and Go Susume! 14:39
jmbrownlee333 Do we know when we are likely to get another crack at the chaperone? 14:40
Skippysk8sirc are there any other factors we should consider when looking at this protein? 14:40
betahelix that has always been our goal... hopefully when the rest of the crystallographic community sees how powerful Foldit players are they will send us many real-world puzzles! 14:41
@TimovdL Like what kind of binding? 14:41
horowsah in terms of things to think about with these proteins: I don't want to bias you all too much. I know that sounds like a copout, but I want to see what you think! 14:42
Brow42IRC This might need a push to make large puzzles run better on foldit. Also, do you have data on player performence with high and low resolution ED? Perhapse players can play with downsampled and smoothed ED. 14:43
@TimovdL Or what our scripts can do 14:43
brunokestemont or both 14:43
horowsah we know binding is quite tight between the two proteins 14:43
horowsah nanomolar dissocation constants 14:43
brunokestemont Should the mobile part be stable enough to go alone in water? (like symmetrics) 14:45
@TimovdL What is that? 14:45
@TimovdL In foldit terms? 14:46
@TimovdL Binding or hydrophobic burying? 14:46
horowsah not sure if i have brunokesetemont's question right- but the mobile part in solution isn't wondefully stable 14:47
jeff101 the ROSIE server has a protocol called docking2. One of its outputs is Isc, a score for how well 2 proteins bind to each other. 14:47
jeff101 ROSIE is based on Rosetta like Foldit is 14:49
spvincent The University of michigan challenge was mentioned earleir 14:49
jeff101 yes 14:49
horowsah the way i think about dissociation constants is that when something is in the nanomolar range, they are bound so often as to be thought of as almost irreversible 14:49
horowsah not quite, but close 14:50
spvincent was there ever a report on the the results? 14:50
brunokestemont ok then we can even imagine a loop that navigates inside of the fixed part? 14:51
brunokestemont a long loop 14:52
horowsah so on the current puzzle, so long as there's not too many clashes, I could imagine it's possible that a loop could thread its way between some of the parts of the fixed parts 14:52
betahelix @spvincent we hope to have the report out soon! 14:53
spvincent tx 14:53
brunokestemont ok, may be trying to bind it with bands ... 14:53
Skippysk8sirc is goal to get different parts that bind tightly, or one good tight binding part and several folds for rest of chaparone protein? or we don't know? 14:55
horowsah basically, all of the above 14:55
Skippysk8sirc :) 14:55
horowsah it's hard to say which combination is going to work the best 14:55
brunokestemont I got a lot of points starting with a simple shake all. Were the starting position sidechains not inside enough of the ED ?(this was not my impression at first look) 14:56
horowsah that's not terribly surprising. some of the flexible regions in the otherwise fixed chaperone were not wonderfully put in the ED, and it so it might be possible to improve those parts simply by shaking 14:57
horowsah i have to head out in a few minutes... any final questions before I go catch my bus? 14:59
inkycatz We appreciate you stopping by! :) 15:00
brunokestemont hopefully we'll be able to load previous solutions from puzzle 1221: none of them was finished because of the crashes 15:00
Skippysk8sirc thanks 15:00
@TimovdL Tx for lots of info 15:00
brunokestemont thanks 15:00
jeff101 thanks for coming. I enjoy these chats. 15:00
BletchleyPark what is this structure for ? 15:01
inkycatz also thanks to betahelix and bkoep from dropping in, and jflat for logging of course 15:01
betahelix and thank you for organizing, inkycatz :-) 15:01
horowsah very generally, the structure is to try and help us understand how chaperones interact with proteins, which are dynamic and non-specific 15:01
horowsah thanks for the great questions, all! 15:01
Skippysk8sirc yes. We do look forward to feedback... 15:02
BletchleyPark ok, bye 15:02
Marktoo bye and thanks 15:02
horowsah I'm going to head out now, hopefully we will do this again once we get the puzzle re-posted 15:02
inkycatz Thanks for coming everyone! :) 15:02
betahelix thanks again! Bye everyone, keep up the great folding! 15:02

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

User login
Download links:
  Windows    OSX    Linux  
Windows
(7/8/10)
OSX
(10.12 or later)
Linux
(64-bit)

Are you new to Foldit? Click here.

Are you a student? Click here.

Are you an educator? Click here.
Social Media


Search
Only search fold.it
Other Games: Mozak
Recommend Foldit
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