Developer Chat

inkycatz It’s officially time for chat to start, so let’s get rolling. 13:58
inkycatz I’d like to welcome everyone to our much belated sci chat - we had a good reason 13:58
inkycatz which i’m sure you’ve all checked out our latest published efforts 13:58
inkycatz Most of our questions today are for and about that but we do have some others 13:58
inkycatz That being said if we can keep the interruptions and followups to a minimum, you know I’ll appreciate that 13:59
inkycatz and so will everyone who posted questions :) 13:59
inkycatz Awesome reminders. 13:59
inkycatz If you are here and did not fill out the demographic survey I assure you it’s important 13:59
inkycatz 13:59
inkycatz Props and thanks to everyone who’s participated in our electron density forum discussion 13:59
inkycatz It’s been a big help 13:59
inkycatz And now on with the chat. 13:59
inkycatz First question! 13:59
inkycatz The Nature communication was published online. Will it ever be published in hard copy? 13:59
bkoep YES, the ED forum was full of great suggestions! 13:59
ScottH howdy, everyone- on the topic of nature communications 14:00
ScottH it is only online 14:00
ScottH i guess they do it to keep costs down 14:00
inkycatz Awww. 14:00
inkycatz Our grand digital future. Next up: 14:00
inkycatz Is the structure published in the PDB (pdb 5KCI) taken directly from a foldit solution, or was additional processing applied? What was the Rosetta score of the foldit solution that was selected for publication in the PDB? 14:00
ScottH so it did undergo a bit of processing 14:01
ScottH basically, we trimmed away parts that were really disordered, since that currently can't be done in Foldit 14:01
ScottH and then we ran it through an automated refinement procedure so that it could be compared with what the other competitors were doing 14:02
ScottH it changed a few bonds and angles very slightly 14:02
ScottH but that's it 14:02
inkycatz What was the Rosetta score of the foldit solution that was selected for publication in the PDB? 14:03
ScottH hmm, i should see if I have that lying around here, I can find it soon enough... 14:04
inkycatz We can always post a follow up in the question thread if it’s not right there :) 14:04
bkoep Hold on a second, I can dig up that number... 14:04
inkycatz Ok :D 14:04
tokens What does disordered actually mean? 14:05
inkycatz (that’s a good question) 14:05
ScottH it is a good question 14:05
ScottH for crystallographers, it basically means any part of the protein where the electron density is too weak to draw conclusions about what's there 14:06
ScottH this is often due to protein motion 14:06
frood2IRC so it’s like a ‘blurred’ area? 14:06
ScottH so there's different levels of disorder, it can be blurred 14:07
bkoep (I believe the best-quality Foldit model had a Rosetta score of -961 REU, and a Foldit score of 17608) 14:07
ScottH it could also be something that is so dynamic that it's kind of like spaghetti flying all over the place 14:07
inkycatz Did any results from Puzzle 1193 go into the Nature paper? 14:08
ScottH no, that didn't end up getting included 14:08
inkycatz next q: Is replication not a concern for the U Mich challenge result? 14:08
ScottH hmm, replication is an interesting topic 14:09
ScottH so in crystallography, unlike most scientific ventures, you actually only have to get it right once 14:09
ScottH the validation stats are usually good enough to tell you if it's bs or not 14:10
ScottH and so you don't usually worry about getting it again in an independent experiment 14:10
ScottH does this answer the question? 14:10
inkycatz I think so. 14:10
jeff101 did different groups get solutions similar to the best one? 14:10
ScottH ah, yes there were quite a few 14:10
inkycatz How did you sample the pictures for the movie? Do you have a track of all the moves of all the solutions? 14:10
free_radical no 14:11
free_radical back :) 14:11
inkycatz wb free_radical 14:11
free_radical I weill answer on behalf of the person who did that work 14:11
free_radical a structure is saved every two minutes, unless you specifically save the structure 14:12
free_radical so the movie was reconstructed from every two minutes 14:12
free_radical unless, of course, you saved the structure, then more data was added 14:12
bkoep Re: replication, I think the scientific community will be very curious to see if Foldit players can successfully interpret other ED datasets 14:13
bkoep So we definitely hope to keep developing ED tools and running more ED puzzles in Foldit 14:13
frood2IRC :D 14:13
inkycatz What is the share of hand/tools for Trained crystallographers? 14:14
ScottH okay, i think i need some clarification on that one 14:14
inkycatz Hey Bruno, you around? 14:15
inkycatz It’s your question :D 14:15
ScottH the trained crystallographers were using Coot, which has tools that are somewhat similar to foldit 14:15
brunokestemont hi, figure 3 14:15
bkoep I think, the question is addressing the amount of manual model-building done by crystallographers 14:15
inkycatz yes that seems like a good approach to that question 14:15
gitwutirc Maybe he meant what sort of tools were available to them. 14:15
brunokestemont rigid body, tweak and pull = hand 14:16
bkoep For context, in Supp. Figure 3 of the paper we break down the types of actions used by Foldit players 14:16
ScottH ah, so Coot is rather primitive in the tools it has 14:16
inkycatz (there you are bruno, thanks for chiming in to clarify and also the question itself!) 14:16
brunokestemont other moves are optimization algorythms 14:16
ScottH it really has just the ability to drag stuff around and then do some local optimizations 14:17
ScottH and making helices on command 14:17
ScottH much less sophisticated than the tools in Foldit, really 14:17
ScottH so the trained crystallographers had that disadvantage 14:17
ScottH as did the students 14:17
brunokestemont ok then they use more "hand" than we do 14:17
frood2IRC depends on how a player here plays 14:18
ScottH basically yes 14:18
bkoep Definitely—Coot users have no Wiggle, no Shake, no scripting 14:18
brunokestemont they should 14:18
inkycatz did that answer your question, bruno? :) 14:19
brunokestemont yes thks 14:19
ScottH so i kinda know the main Coot developer. I'll be interested to see if he wants to make it more like Foldit 14:19
inkycatz do keep us posted there! 14:19
inkycatz that’d be…interesting 14:19
inkycatz next question! 14:19
inkycatz It’s a big one: 14:19
inkycatz YPL067C being a yeast protein, what do you expect with this discovery? 14:19
inkycatz -could researchers then find similar Human HTCs that regulate DTA? 14:19
inkycatz -is YPL067C a toxical and/or a potential drug? 14:19
ScottH so it's a bit early to speculate on it too much 14:20
ScottH it's an interesting lead for studying prion diseases and neurodegenerative diseases 14:20
ScottH and maybe helping develop therapeutics 14:20
ScottH but we're a long way from knowing any of that for certain 14:20
ScottH so far, it doesn't appear that humans have one of these proteins 14:21
ScottH maybe it evolved to handle something similar in fungi? 14:21
ScottH too early to tell 14:21
Susume2 do fungi have a-synuclein? 14:21
ScottH i don't think so, but they have lots of prions that also form amyloids 14:22
ScottH so i think there's more evolutionary pressure to develop good defense systems when they go awry 14:22
ScottH since most amyloid diseases develop past parenting age for most people, there's less evolutionary pressure to have proteins to handle it 14:23
inkycatz OK last paper question we have posted: What about the full structure? (incl. unordered) 14:23
inkycatz The Foldit players better matched the density after removing the unordered segments. What about refining these unordered segments? 14:23
inkycatz 1) Aren't they important for the full function of the protein? 14:23
inkycatz 2) Are these segments part of the (before competition) published structure? and where them there correct? 14:23
ScottH good questions 14:24
ScottH so are those segments important: we're not sure yet, we're in the process of testing them now 14:24
ScottH were they part of the "before competition" structure: in part, yes 14:24
ScottH but mostly no 14:24
ScottH mostly those disordered regions are not in the published version 14:25
ScottH i have ideas for how Foldit players can help with disordered sections of proteins, but it still requires some more tweaking 14:25
ScottH not ready for prime time just yet 14:25
inkycatz Feel free to ramble out a blog for some feedback from the community anytime 14:26
inkycatz Just saying 14:26
ScottH ha, yes that might happen sometime 14:26
inkycatz OK time for the not paper related questions! All our scientists feel free to chime on in. :) 14:26
inkycatz Have foldit designs gotten more realistic since the Ideal Loop filter was introduced? Are any being produced for lab testing? What problems remain with the designs you are seeing now? 14:26
ScottH oh wait, just looked it up 14:27
ScottH synuclein does occur in fungi 14:27
ScottH sorry, back to the new questions 14:27
inkycatz that’s good to know, I love learning new things in science chat :D 14:27
bkoep Yes, Foldit designs have definitely gotten more realistic with ideal loops 14:27
brunokestemont :) 14:28
bkoep The protein backbones are looking much better, and we are generally seeing more of them perform well in our downstream analysis 14:28
bkoep We have gotten some more of them into the lab, and a couple proteins look promising 14:29
bkoep High resolution data (i.e. crystal structures) have proven elusive, but we're still working on that 14:29
bkoep I realize it's been a while since we had a blogpost with new updates—I apologize for the lack of communication. 14:30
bkoep We are still seeing isolated cases of problematic backbones—our best hypothesis is that these are coming from the Rebuild tool 14:30
inkycatz ready for more questions? 14:31
bkoep We also hope to roll out some new tools very soon that will make it easier to construct "ideal loops" 14:32
brunokestemont Q: ED of the paper was medium wiggle power, end refining went away from native; is High wiggle power gives points but seem to escape from native. Is it really usefull? 14:32
bkoep @bruno, that's a good question, about the use of high wiggle power 14:33
alcor29 I ran a test  and used only rebuild tool and a following wiggle. Led to good Rama.So is it possible that it is not the tool but the scripts that are causeing the deviations? 14:34
bkoep There was a graduate student in the Baker Lab a few years ago who showed that a similar type of "high power" minimization improves discrimination between native and decoy models 14:34
bkoep In Foldit, however, we haven't seen that high power wiggle leads to more accurate models 14:34
inkycatz new questions: Recent denovo puzzles look like they were designed in foldit (possibly by foldit players). Can you confirm this? Also can you comment on how the designs were chosen and on what you hope to learn from posting these puzzles? 14:35
bkoep @alcor—it's possible that the scripts are exacerbating problems by repeatedly introducing many small imperfections, or something like that... 14:36
bkoep But, the types of backbone "imperfections" that seem to be problematic can still have good rama scores 14:37
bkoep Re: denovo puzzles, yes, we've used Foldit player designs as targets for several recent De-novo Freestyle puzzles 14:38
bkoep In general, these are Foldit designs that look especially interesting to us, or that stand out in some way during downstream analysis 14:39
bkoep Often, the are protein designs that look good in manual inspection, but that aren't sampled well by our Rosetta@home analysis 14:39
inkycatz Interesting stuff! 14:40
bkoep Since Rosetta can't seem to find the designed fold, we're interested to see if Foldit players are able to find the designed fold 14:40
inkycatz And here’s our last one from the pile today: Are helices made by ideal SS really ideal? 14:41
bkoep Excellent question! 14:42
bkoep Susume made a great observation in the forums about Foldit helices that had completely escaped our notice 14:43
bkoep 14:43
Susume2 (tokens clued me in to it) 14:44
inkycatz (everyone caught up on that post?) 14:44
bkoep In short, we hadn't paid a whole lot of attention to the "ideal helices" and "ideal sheets" in Foldit. 14:45
TomTaylor5 Tweaking also changes the helix shape. What about that? 14:45
bkoep The "ideal" phi/psi torsions that we're currently using are not optimized with any degree of precision, because the minimizer (Wiggle) is very very good at that kind of optimization 14:45
bkoep But there are definitely ways in which these imprecise secondary structures can lead to deep local minima in the energy landscape that Wiggle is unable to escape 14:47
bkoep Basically, it never occurred to anyone that it would be a problem to use imprecise helix/sheet definitions, but that could very well be the case 14:49
jeff101 do the ideal phi/psi torsions depend on amino acid types and nearby amino acids? like proline vs glycine vs alanine 14:50
bkoep So yes, the ideal helix/sheet tools are due for an overhaul, and hopefully we can get that out to all of you very soon 14:50
bkoep @jeff101, yes different amino acids have different preferences for phi/psi torsions 14:51
bkoep But in the case that those preferences do not align with the phi/psi of a stable helix, you simply don't find those residues in helices 14:52
bkoep Or else those residues can be used to construct "non-ideal" sheets and helices 14:53
bkoep (e.g. helix kinks, and beta bulges) 14:53
bkoep Also, I do want to take this opportunity to thank everyone for the excellent ED discussion in the forums 14:55
bkoep 14:55
inkycatz Yes, feedback is always very important! 14:55
inkycatz And it has been a great discussion. 14:56
inkycatz BTW we are almost out of official time but I’d like to say thanks for coming to all of you, and also to our scientists today :) (and jflat for logging as always) 14:56
ScottH thanks everyone, as always this is fun 14:56
jeff101 thanks again for having these chats 14:57
inkycatz Yes we’ll try to get back on a regular schedule 14:57
inkycatz we just wanted to wait til the paper was out for this one! :D 14:57

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.
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