Aflatoxin Round 3

Case number:699969-2004460
Opened by:jeff101
Opened on:Tuesday, November 14, 2017 - 21:38
Last modified:Wednesday, November 15, 2017 - 23:01

Assuming there will be another sequel
to the Aflatoxin Puzzles 1440 & 1445,
I would like to request the following:

(1) Let us continue to input solutions
from Puzzles 1440 & 1445 into the latest
Puzzles in the series. I know I have some
personal favorites that I would like to
keep improving.

(2) Include as optional starting structures
the Foldit Team's favorite structures from
previous rounds. This would give us a better
idea what you are looking for. It might also
let different teams work their magic on other
teams' best solutions. Maybe members of one
team will find things that members of other
teams missed. Maybe special Recipes only
available in certain teams can work their
magic on other teams' solutions. It could
also balance out the playing field so that
winners of previous rounds don't have such
a great advantage for subsequent rounds.


(Tue, 11/14/2017 - 21:38  |  8 comments)

toshiue's picture
User offline. Last seen 20 min 38 sec ago. Offline
Joined: 01/31/2016
Groups: Go Science

Agreed. Haven't half the time I'd like in order to implement what I think I've learned I should be doing here. As to some variant on the "all hands" scenario, on such a major thorny issue as aflatoxin, I think out priority is to bring all possible talent to the table and set aside score issues for the duration.

Susume's picture
User offline. Last seen 3 hours 46 min ago. Offline
Joined: 10/02/2011

Strongly agree - please give us several promising solutions from past puzzles as starting points. If you want variety, put a filter that requires a certain score increase or certain amount of difference in backbone or AAs.

Also any feedback you can give about what you want to see more of (or less of) would be very helpful.

Joined: 06/06/2013
Groups: Gargleblasters

Adding segments didn't help me at all. my first puzzle is still my best score. I seem to have no idea how to work this, as I can increase bonds to the peptide and watch my score drop. never mind adding segments. I just gave up trying to even get anything bigger up to old puzzle score
Please idealize before listing again. It is a big puzzle for a little machine, and the backbone is awful. Seems a waste of time to have to do it again each time if want to restart. Also, would love to see a way to flag AAs we added so we can consider deleting them and trying something else
I like Toshiue's all hands on deck idea....

toshiue's picture
User offline. Last seen 20 min 38 sec ago. Offline
Joined: 01/31/2016
Groups: Go Science

Skippy brings up another very good point (they were all very good points) on the “all hands” issue, that of material resources. In those instances where “all hands” might be implemented, we need an equitable protocol by which those of us with extra clients can opt to share such with those having more ideas than CPUs. Six good ideas and a dual core PC is just a waste of talent.

jeff101's picture
User offline. Last seen 9 hours 2 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

It would help if when you Tab on a residue
(like #303 in 1445, the Aflatoxin), it
would list the # of hydrogen bonds that
residue is involved in. The Bonding Score
does somewhat reflect this, but there is
much variation, as in the examples below:

 3 hbonds -> Bonding  8.7 to 16.2
 4 hbonds -> Bonding 22.0 to 22.1
 5 hbonds -> Bonding 24.2 to 29.7
 6 hbonds -> Bonding 29.9 to 37.4
 7 hbonds -> Bonding 34.9 to 39.3
 8 hbonds -> Bonding 32.2 to 43.7
 9 hbonds -> Bonding 40.8
10 hbonds -> Bonding 41.1

I know you can show hydrogen bonds to
non-protein residues like the aflatoxin,
but with bands in place and other 3D
obstacles, it is a pain to count them all.

toshiue's picture
User offline. Last seen 20 min 39 sec ago. Offline
Joined: 01/31/2016
Groups: Go Science

Seriously agree with Jeff's point on hydrogen bond listing issue. That lack of such was probably the greatest unproductive time consumer for me on 1445.

Joined: 05/19/2017
Groups: None

I betcha folks that they're gonna work through some score filters to add to the next rounds but only after some time. If I had to make an educated guess, score buffs to H-bonding, especially the oxygens, will probably turn up. This is, after all, the transition state that we're binding; if we can get the enzyme to stabilize the transition state it makes the activation energy lower and the cleavage easier to do. I would also educated guess that filling space around the molecule could receive score buffs

Give it time though, I reckon that they've gotta sift through a whole bunch of noisy data first with sorting scripts before they even start wet lab stuff, much less reporting on what's working.

If you guys are curious I prioritized H-bonding in my solutions so if you check my Twitter picture posts frequently you'll see some screenshots in there of me manhandling some polar residues with Freezing/ZLB's to survive the recipe refinement storms. I can't say with certainty that they're the "good science" solutions, but it might inspire something, so have a look! I'm happy to share. Here's a link to the latest one I have from Round 2
This solution did not have as high of a score as my other one from Round 1 (I ended up letting mutate and recipes eat some of the less important looking H-bonds), but in both cases, I did go out of my way to preserve sub-optimally scoring bonding residues since I've got a hunch that we need those the most to hold this rascal in place.

bkoep's picture
User offline. Last seen 7 hours 7 min ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff

This is great feedback!

I'm sorry that we won't be able to meet these requests in the Round 3 puzzle, but I'm hopeful that we'll be able to in a later round.

In the Round 3 puzzle, we wanted to try the ligand in a different orientation within the binding pocket (there are two orientations of aflatoxin that we think could be amenable to the desired reaction). Unfortunately, this means that solutions from the previous rounds are not compatible with the Round 3 puzzle.

The Siegel Lab has begun their analysis of Round 1 solutions, and will likely start digging into Round 2 later this week. I think it's a great idea to revisit some of the more promising designs in a followup puzzle, and we've asked them to keep an eye out for solutions that they would like to see developed further.


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, Boehringer Ingelheim, RosettaCommons