Closed since over 10 years ago
Intermediate Overall DesignThis is a basic 65 residue design puzzle without any Fragment Filter. See the puzzle comments for filter details. The Baker Lab will run folding predictions on your solutions for this puzzle, and those that perform well will be synthesized in the lab. Remember, you can use the Upload for Scientists button for up to 5 designs that you want us to look at, even if they are not the best-scoring solutions!
100 pts.
10,282
Residue IE Score: Monitors that all PHE, TYR, and TRP residues are scoring well.
Core Existence: Ensures that at least 30% of the residues are buried in the core of your design.
Secondary Structure: Checks that no more than 50% of residues are in helices; penalties are incurred if more than 50% of residues form helices.
Secondary Structure Design: Penalizes all CYS residues. Penalizes GLY, ALA, and PRO residues in helices and sheets.
Did any of the previous 65 residue designs make it to the real world ?
Which (type of) designs were tested, if any ?
I'm sorry we haven't posted a design blog post recently. We haven't directly tested many designs since introducing the Secondary Structure filter, but that does not mean the project has stalled!
Foldit players have been consistently designing with some really cool α/β folds, but for some reason they are less likely to pass our computational analysis, compared with Baker Lab designs. Recently, all of our effort has been focused on figuring out exactly what the problem is and how to address it. We should have some updates to share soon!
I've the impression that the filters interfer with the score in many unintended ways.
This is illustrated when we try to wiggle with bands enabled: there is no unique solution and Foldit turns around many attempts without stabilizing on an optimal solution.
When we try to mutate with bonus or penalties on specific AAs, it's clear that the mutate function doesn't take these constraints into account: we get a lot of "bad" mutations that we have to correct afterwards manually. My strategy is to mutate "manually" or "one by one with scripts similar to Mutate No Wiggle. Doing this, I loose the advantage of the embedded mutate tool. May be my mutations aren't optimal, but they are rewarded by (artificial) bonus.
For SS, the same problem occurs: we get a big penalty when there are too many helices. But this "artificial" penalty interferes with the embedded score function.
For Core filter, it's even more complicated: the filter interfers all the time with wiggle. I wonder if the wiggle function is still "optimizing" anything in this situation.
I wonder if it would not be better to avoid many interference from filters in the score function, "simply" by changing some filters to conditions and by eliminating some AAs from the internal list of possible AAs during the mutate action.
Replacing SS filter by SS Conditions: we'd read "SS condition not met: too much helices". You could ask us, since the beginning, to build a design with "maximum x residues in helices". Or even more precise: "60 residues HHSSSS design puzzle" would ask us a design with 2 helices and 4 sheets. Or you give us a rigid SS of this kind with embedded allowed AAs. Our freedom is then limited to the PS and TS, with the advantage that less filters interfere with the score functions.
Or even an additional step that would be a kind of De Novo on our own design: we may start from our designs from former puzzle, but no mutation is available anymore, say the latest 2 days or in a new puzzle round.
Not sure that the players would appreciate these "rigid" constraints, but we could gain on "slow filter" burden reduction.
