Developer Chat

@inkycatz WELCOME to our first science chat of 2016 :D Feels like it’s been forever, really. But that’s the holidays for ya. 12:31
@inkycatz Some quick housekeeping 12:31
@inkycatz First of all - question priority goes to those who posted ahead of time 12:31
@inkycatz (for any newcomers) 12:31
@inkycatz HEY! 12:32
@inkycatz bkoep :) 12:32
@inkycatz (we were just doing the general housekeeping stuff) 12:32
@inkycatz First of all, chats are logged, so if you have to go don’t worry we post the log afterwards. 12:32
@inkycatz Obligatory social media announcement - you can find us on Facebook 12:32
@inkycatz https://www.facebook.com/folditgame/ 12:32
@inkycatz And Twitter: @foldit 12:32
bkoep Hello! 12:32
@inkycatz This helps in case of any server issues that may or may not be going on. 12:32
@inkycatz By the way, thanks jflat for the logging! :) 12:32
@inkycatz Its key to the process ;) 12:32
@inkycatz I’d love to start with the snowflake design puzzle honors! 12:33
@inkycatz spvincent, 01010011111, actiasluna, smilingone, franse and one more community member (who hasn’t gotten back to us yet to give us permission to identify them)! 12:33
@inkycatz Excellent work, all of you 12:33
@inkycatz Look for that blog post sometime later today with pictures and who on the team picked what as favorites :D 12:33
@inkycatz Sorry, it’s been a busy morning and I didn’t get to posting it before. 12:33
@inkycatz To start off today 12:34
@inkycatz free_radical and the ongoing saga of drug design puzzles. Take it away! Oh, and could you answer our first question of the day, “what are the grids for?” :) (appreciated) 12:34
free_radical sure, first an update 12:34
free_radical A lot of people wanted to use the drug design tools with the selection interface, so, as per request, I ported all the new tools over to the selection interface! 12:34
BletchleyPark does it still work in the original interface ? 12:35
free_radical I am also in the process of adding LUA functionality to the tools so that everyone can use the cookbook to come up with new recipes! 12:35
free_radical yes, the original interface is still there 12:35
free_radical I will update them together as new tools come online 12:36
free_radical I recieved a lot of great feedback from the devprev puzzle and I have slowly been implementing everyones request, to the best that I can 12:36
free_radical I want to thank everyone for all your suggestions. They were great and are going to help make the puzzles nice for everyone 12:37
free_radical So, lets talk about the grids, unless anyone has any more questions 12:37
free_radical When scientists design drugs computationally, they useally have information on the target they are working on. This information comes from the protein, or other drugs that they have developed 12:38
free_radical This "information" is usually called pharmacophore maps or fingerprints 12:38
free_radical It is hard to actually display this information in a meaningful way. It can be displayed on the protein side or on the ligand side 12:39
free_radical The grids that were shown were my attempt to show this information 12:40
free_radical In the most simple form, I was trying to highlight the atoms that form hydrogen bonds 12:40
@inkycatz That makes sense. 12:41
free_radical in more complex problems, the information shown can be parts of the protein that like polar atoms, or have an affinity for charged atoms 12:41
free_radical or hydrophobic atoms 12:41
free_radical So, the grids were simply an attempt to show information on where you should target your designs 12:41
free_radical now, this was a first implementation and I hope to have an improved version out soon 12:42
free_radical just as an example 12:43
Skippysk8sirc when it is working, can we have a video on drug design for dummies please? 12:43
@inkycatz :D 12:43
free_radical here is another representation of the maps 12:43
free_radical http://www.webmedcentral.com/articlefiles/9568c41349dd0efe57c1ed67d3234a56.png 12:43
free_radical I guess links dont work? 12:43
Skippysk8sirc worked 12:43
free_radical oh nevermind, cool 12:43
@inkycatz seemed to work for me :) 12:43
@Susume2 https://tinyurl.com/5p2yvd 12:43
frood2IRC much better 12:43
smilingone Links work for those of us on external irc 12:43
@inkycatz (thanks susume) 12:43
free_radical thanks :) 12:43
free_radical so, I am far from an artist, but I try :) 12:43
frood2IRC :) 12:44
free_radical drug design has a lot of complex information 12:44
free_radical and trying to distill this information down into components that are easy to understand is difficult, but we are getting there :) 12:44
free_radical and I am always open for suggestions 12:44
@Susume2 the spheres make more sense to me visually than the grids did 12:45
free_radical cool, I think I can get something like that up and running 12:45
free_radical I know that the grids caused a lot of confusion, so lets try something else 12:46
free_radical To answer the question about a video, is there anything particular you would like? 12:46
jeff101 start a puzzle from scratch and say what you look for 12:47
free_radical I think maybe a video on the process of how drugs are created? 12:47
jeff101 then start doing it 12:47
free_radical and where Foldit fits into the process? 12:47
BletchleyPark explaining the strategy you would recommend and explain the goal for the exercise 12:47
Skippysk8sirc basic goal, use of each tool so we can provide the answer you want. the last time I tried ages ago I couldn't even add segments 12:47
free_radical cool, more work for my undergrad :) 12:48
@inkycatz Sweet 12:48
jeff101 if it takes a long time, skip some intermediate steps and just show the key steps 12:48
Skippysk8sirc ;) 12:48
frood2IRC LOL 12:48
@inkycatz And more content for our youtube channel! 12:48
Skippysk8sirc only on Jeff's version 12:48
@inkycatz hahaha 12:48
free_radical we do have a nice tutorial puzzle 12:48
free_radical where the dialog walks you through the tools and what to look for 12:49
free_radical it mimics a nature paper and how they went about designing the small molecule 12:49
free_radical with the next release of drug design, that tutorial should be there 12:49
@inkycatz We appreciate the update! 12:50
free_radical de nada 12:50
@inkycatz I think we’re ready to roll out some of these great posted questions. First topic, hbonds! 12:51
@inkycatz “With regards to h-bond networks, should we be putting the large hydrophobes at the surface (or where the surface meets the interior) so they can bond with water?” 12:51
@Susume2 not just in hbond puzzles, but in all puzzles 12:51
@inkycatz (thanks for the clarification there :) 12:52
bkoep Yes, Susume had a great question about large hydrophobic residues that contain one or more polar residues 12:52
bkoep Namely, TYR and TRP 12:52
frood2IRC and paricularly ryp and tyr 12:52
frood2IRC grr - my typing - sorry 12:52
bkoep Conventionally, we consider these residues to be hydrophobic 12:53
bkoep The sidechains are dominated by hydrocarbons that cannot form hydrogen bonds, and are much more stable buried in the core of the protein 12:53
bkoep However, TYR and TRP both contain a polar atom that very much likes to make a hydrogen bond 12:54
bkoep So, it really is best to place these residues in such a way that you bury as much surface as possible, while still satisfying the hydrogen bond 12:55
vmulligan In natural proteins, we see that polar atom either makking hydrogen bonds in the core (part of hydrogen bond networks) or "snorkeling" -- interacting with solvent while the more hydrophobic parts of the sidechain stay buried. 12:55
frood2IRC does this also apply to histidine  - inversely? 12:56
bkoep I don't think this is so much of a problem for histidine 12:57
frood2IRC kk - just wondered if it could do well hidden 12:58
frood2IRC thx 12:58
bkoep Histidine is often semi-buried in enzyme active sites, but it is also found on the surface of proteins as well 12:58
vmulligan Yes, histidine tends to be pretty polar.  Most of the side-chain can form hydrogen bonds, so there's not a lot of hydrophobic surface area there to be buried. 12:59
@Susume2 some kind of marker to tell us if the polar can currently reach the water would help us with building snorkels 12:59
jeff101 view as spheres with all sc's and hydrogens shown? 13:01
bkoep Hmm, Susume, that's a good point. 13:01
jeff101 color by AA? 13:01
bkoep For the most part, you can probably eye-ball it pretty effectively 13:01
BletchleyPark isosurface view ? 13:01
bkoep Yes, isosurface view would be a good way to test whether you can "see" the polar atom from the surface 13:01
@Susume2 I'll try that 13:02
@inkycatz next question time? 13:02
Skippysk8sirc thanks for snorkel analogy vmulligan 13:03
@inkycatz (yeah that was pretty helpful) 13:03
@inkycatz “Can we expect to see more ED puzzles in the near future?” 13:03
bkoep We hope so! 13:03
bkoep It depends on when fresh data becomes available. 13:04
@inkycatz There is a great thread with a lot of questions on this but I picked out one follow up: On Puzzle 1152, I have seen the "Density" subscore be nonzero, sometimes positive and sometimes negative, even when the protein is far away from the ED cloud. Is the exact same ED cloud tiled repeatedly through 3D space but only shown for one small region of 3D space? What do negative "Density" subscores mean? 13:04
jeff101 yes, could you only have the density subscores be nonzero in a region near where the density is shown? this would give us the option of folding the protein outside the density and then later moving it into the density 13:05
bkoep Yes, the ED cloud is tiled indefinitely in all directions 13:06
frood2IRC the current tutorial flies in the face of this 13:06
@inkycatz (let’s save the editorial and stick to questions, please) 13:07
bkoep When we curate the density map (as in puzzle 1152), we usually "pad" the ED cloud with some amount of zero density so that neighboring tiles aren't too close to one another 13:07
BletchleyPark if you had virtual symmetry copies included in the model, would that help to match the density from a scoring perspective ? 13:07
bkoep In Puzzle 1152, we may not have provided enough padding, so you might have experienced nonzero density close to the visible ED cloud 13:08
frood2IRC would it be very hard to restrict density scoring to the rendered are only? 13:09
frood2IRC *area 13:09
jeff101 if the visible ED cloud occupies a cube w/sides of length L, why not let the density be zero outside a cube w/sides of length 3L or 5L? 13:09
jmbrownlee333 Are we seeing just  one asymmetric unit then? 13:09
bkoep @BletchleyPark, yes, ideally we would like to work with a full unit cell (uncurated density), in which case the symmetry function would probably help to build a good model 13:10
bkoep Unfortunately, rendering the full unit cell and multiple protein copies would slow down Foldit significantly. This is why we try to limit the amount of density you have to work with. 13:10
BletchleyPark ok, thanks 13:11
bkoep @frood, I think that's something we could do, if neighboring units are a real problem for players 13:12
frood2IRC I’m just trying to find ways to avoid current misconception - many players believe that folding outside this area is not affested 13:13
jeff101 I'd just like to be able to begin folding w/o interference from the ED 13:13
jeff101 and then later move a good fold into the ED 13:13
bkoep @jmbrownlee333, yes, we try to limit the ED cloud to one asymmetric unit when we can 13:13
frood2IRC thx jeff - that;s my point 13:13
jeff101 ideally, the protein should know how to fold itself correctly without the ED there 13:13
bkoep Of course, we can only do this when we have a pretty good idea of the where the model is placed in the unit cell 13:14
jeff101 and the align protein to the ED tool is terrible! it usually lowers the score 13:14
frood2IRC ans then we have the edges - always the dodgy bit  :) 13:15
@inkycatz (again, let’s keep the editorial minimal, we have plenty of questions left here :) 13:15
frood2IRC sorry 13:15
@inkycatz Let’s move on to one of those new questions: What is the source of the current de-novo designs? 13:15
bkoep A while ago, we tried an ED puzzle in two rounds: in the first round the density was visible but did not contribute to scoring 13:15
@inkycatz (when bkoep is done naturally :D 13:15
jeff101 if you confine the nonzero density to a cube of length 3L or 5L, it gives you some wiggle room 13:15
bkoep Was that helpful, or would it be much more useful to move into and out of the density at your convenience? 13:16
jeff101 I'd like the option to move in and out of the density 13:16
jmbrownlee333 could the density contribution to score be toglled? 13:17
jmbrownlee333 togglrd? 13:17
jmbrownlee333 toggled? 13:17
bkoep @jmbrownlee333, that's an interesting feature that we've considered implementing (full control of individual score terms, a la "Clashing Importance") 13:18
bkoep I'm not sure if we'll be able to do that effectively 13:19
bkoep Moving forward, I think bertro had posted a question about the recent de-novo freestyle puzzles 13:20
@inkycatz yep! 13:20
bkoep Some of the recent De-novo Freestyle targets have indeed been sequences of Foldit player designs 13:21
@inkycatz Oh cool 13:22
@inkycatz This is about residues scoring “too low”. Bruno asks: I just wonder why some residues should not score "too low" in filters, and how much is this "too low". Is that a general rule, including for de-novo? 13:22
@inkycatz Good timing bruno, we just got to your question about filters and scoring and whatnot 13:23
bkoep Briefly, about the de-novo freestyle puzzles: I just want to say that these sequences are not necessarily successful Foldit designs 13:24
bkoep Usually, these are designs that Rosetta structure prediction is unable to recapitulate—possibly due to its limited sampling of structure space. We simply want to compare Rosetta sampling with Foldit player sampling for these designs. 13:25
bkoep About Bruno's question: 13:26
jmbrownlee333 are they then compared to reality? 13:26
free_radical we are in the process of examining that 13:27
free_radical :) 13:27
jmbrownlee333 :) 13:27
vmulligan Yes, but we have a finite capacity for the time- and labour-intensive experiments that let us solve structures experimentally.  So we have to focus on the best small subset of desigs for experimental validation. 13:27
jmbrownlee333 the should be great NMR targets. right? 13:28
bkoep For a couple of reasons, Rosetta (and Foldit) favors large aromatic residues a little more than it should 13:28
bkoep If we left things up to Rosetta, we would get a lot of designs that are mostly TRP 13:29
frood2IRC as we once did here I seem to remember 13:30
bkoep The Residue IE filter is just one way to cut down on the number of these residues that make it into the design. 13:30
bkoep However, I'll mention that there have been some recent improvements to the Rosetta energy function that we hope to incorporate soon into Foldit 13:31
bkoep Don't worry, these changes will be much less disruptive than the "New Chapter" update, for those of you that remember 13:31
@MikeCassidytoo ;-) 13:31
frood2IRC hehe 13:32
@inkycatz I think we have time for this good question before our team has to get back to the lab, plus you know, any updates or comments they want to add before dashing off: Could you tell us more about some of the goals of the design puzzles? Is one of the goals to find proteins that fold in to shapes that nature hasn't found much use for? 13:32
bkoep The score functions should at least improve the quality of Foldit designs; at best, it may even allow us to phase out some of the filters 13:32
@MikeCassidytoo Good 13:33
frood2IRC that would be great bkoep  :) 13:33
BletchleyPark and: are there any new papers in the making on foldit progress ? 13:33
vmulligan The latest Rosetta changes also will let us pull in some design-friendly Rosetta features that have been added recently, including better control over the packer ("shake") to give more sensible amino acid composition. 13:33
frood2IRC dare I ask - any hews on the U michigan competition? 13:34
@inkycatz Guys 13:34
frood2IRC *news 13:34
@inkycatz Let’s let them catch up the posted question first 13:34
@inkycatz Thank you 13:34
frood2IRC kk 13:34
bkoep Aha, @frood, we're still analyzing the results of the UMich competition. Look for updates in the next few weeks. 13:34
frood2IRC kk - thx 13:34
bkoep About spvincent's question on the goals of Foldit design puzzles 13:35
brunokestemont thks 13:37
bkoep Protein design is still a very difficult problem. As much success as we've had in the Baker lab, it's hard to claim we really understand the rules of protein design until we can formalize them into hard criteria 13:38
bkoep Basically, expert protein designers use many different methods and look at a lot of different things when designing a protein, and we still aren't completely clear on what's important and what isn't 13:40
vmulligan (Got to run -- take care!) 13:40
alcor29 Can you address spvincents question on parallelism in this context? 13:40
@inkycatz (thanks for coming vmulligan!!) 13:40
bkoep Foldit players are an enormous help for this problem, because you can test our "rules" very rigorously 13:41
@inkycatz :D 13:41
bkoep In addition to that, protein design is an open-ended problem, and we've seen that Foldit players are able to conceive of protein designs that would never be generated by our automatic algorithms or by trained biochemists 13:43
bkoep So we think this combination of rigorous testing and inexhaustible ingenuity will allow Foldit players to teach us a lot about the basic science of protein design 13:44
smilingone have you found any of them useful?  the ones from untrained players 13:44
bkoep @smilingone, yes, maybe in a few specific cases, players have given us new ideas on H-bond networks, and some of the helical designs have shown extraordinary stability in the lab 13:46
jeff101 be generous with your praise so that we stay motivated 13:46
brunokestemont :D 13:46
@inkycatz (we thought you all knew you rocked, but we’re happy to say it again) 13:46
bkoep But I think the stronger contribution is more general than that. The volume and diversity of Foldit player designs give us a very useful dataset to learn from. 13:47
bkoep And on that note, I have a bit of unfortunate news... 13:47
@inkycatz I suppose that’s really a matter of perspective, but some folks may not like it sure 13:48
bkoep Since there have already been some stirrings about CASP12, I want to warn players that we will not be participating in CASP12 as we did in CASP10 and CASP11 13:48
@Susume2 :-( 13:49
brunokestemont what abour "unfinished" starting designs (not well evolved) but interesting? 13:49
free_radical I can soften the blow...we should have more drug design puzzles out by then! 13:49
BletchleyPark Why will foldit not participate in casp12 ? 13:49
@inkycatz You’re a ray of sunshine, free_radical 13:50
bkoep IF we participate at all, it will probably be only a handful of targets that are hand-selected as those most likely to benefit from Foldit players 13:50
bkoep There are too many other projects that we want to focus on, at the moment 13:51
brunokestemont ok 13:51
jeff101 will teams get to submit their own solutions for the puzzles you pick? 13:52
bkoep In the past, the summer onslaught of CASP targets has prevented us from running design puzzles and puzzles for more pressing scientific problems 13:52
BletchleyPark Let us know if we can be of help with the other projects 13:52
jeff101 should teams register for CASP12? 13:52
@inkycatz I appreciate our team staying over our standard time, now that I’m looking at the clock :) 13:52
@inkycatz (I should just make chats longer since they just go over anyway and aren’t frequent.) 13:53
bkoep Especially with free_radical's drug design improvements, I expect we'll have plenty of projects to work on 13:53
@inkycatz I trust the team will have plenty of cool stuff to present, puzzle wise :) 13:54
BletchleyPark sounds good, eager to take part. 13:55
@inkycatz I’d like to thank everyone for coming and taking part in chat today, by the way. 13:56
@inkycatz And posting those questions ahead of time - it really helps steer the chat a bit! 13:56
bkoep @jeff101, we'll have more information as summer approaches. If we do run any CASP puzzles, I think we can probably work with Foldit players who would like to register their own teams. 13:56
smilingone can you post links to those snowflakes?  I'm curious to see which of mine caught the eye. 13:56
@inkycatz When we’re done with chat I’ll be working on the snowflake post, yeah 13:57
@inkycatz and post a link here in chat when I post it up :) 13:57
@inkycatz I apologize my morning got away from me in that regard, was hoping to do it ahead of time! 13:57
@inkycatz Thanks for all the great info today, bkoep and free_radical! (and vmulligan, when he reads the log) 13:58
@inkycatz Our next science chat will be in March 13:58
@inkycatz and logs will be posted soon 13:58
brunokestemont thanks team ! 13:58
bkoep Thank you all! 13:58

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
PlayerPuzzleScore
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