Get your questions ready, folks. It’s time for a devchat! Come chat with our developers in #global this Monday 3/25 from 11:00am-12:00pm PDT / 18h00-19h00 GMT. We’re on Pacific Daylight Time here in Seattle (-7 GMT) so don't forget to double check your time zone conversion. A transcript will be available on the site after the chat for those who can't make it.
If you need help getting into #global chat, check out this helpful guide for assistance.
Thanks and we’ll see you there!
Edit: A transcript is now up for those of you who missed Monday's chat. Thanks!
I would propose the following to make design and particularly dimer design a little easier:
Core polar residues are okay if they have sidechain hbonds.
Exposed cysteines be allowed if they form a bridge.
It's really hard or impossible to satisfy core polar and layer design for both the monomer and dimer conformation, although core existence is doable.
The filters make puzzles like 690 very very slow. And because wiggle (and maybe also shake) do not take filter scores in account, wiggle at CI 1 can lose points now, strange scorejumps are there.
Moreover, the absolute best and recent best scores can not cope with the filter on and off switch. This gives problems with restoring the best score and evoing, see also the feedbacks about these kind of problems. If the switch is there to stay provisions should be made for that.
To measure how the filters influence the running speed I did run a timing script on 687 and 690 with filters on and off. (I shared this script with myself so it can be used by the devs). The script changes 10 segments to all possible acids without running wiggle or shake. Running times:
690 filter on 418 seconds, filter off 66 seconds
687 filter on 146 seconds, filter off 36 seconds
Conclusion filter on is more than lineair to the size of the puzzle and make the puzzles more than 4 times slower. That is way too much.
Many are finding that the option to run with filters disabled is not viable, because so often changes which improve the protein with filters off can make it worse when filters are turned back on. I ran a rebuild script overnight with filters disabled, appearing to gain 400 points, but when I turned the filters back on I had actually lost 800 from where I started (i.e. filters reduced my score by 1200). I had to discard all the work of the script and re-run it with filters on. Since then I dare not run without the filters.
Couple more questions for the dev chat:
On puzzles like the catalyst (688b) do you only look at each person's top scoring solution, or will you also look at lower scoring tracks? I often have a lower scoring track that seems more realistic.
Can the mutate function in foldit be modified to respect the filter rules? It still chooses lots of invalid sidechains including large exposed hydrophobes, glycines in helices and sheets, and alanines in the interior. Modifying the mutate function would not stop Rosetta from predicting folds for natural proteins that don't follow the filter rules, since mutate is not used on natural proteins.
