6 replies [Last post]
Sporeo's picture
User offline. Last seen 1 year 14 weeks ago. Offline
Joined: 01/20/2012
Groups: Go Science

What do you do at this point when there appears to be no gains to be made, despite seeing other players flying past you by 100s of points? Is this indication of a major fault in the folding and a sign that it's time to start over completely? Is a small tweak via hand-folding a possible solution, or is it worth making a large adjustment that seems out of place and refolding from a presumed improvement despite a large score drop? Is there a script that works amazingly for these conundrums? How do you evaluate the protein at this point and what's your go-to at these points as I find myself stuck between stubbornness and cluelessness at these times.

jeff101's picture
User offline. Last seen 23 hours 34 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Some ideas:

Being in a group, you could share stuck solutions and ask teammates for help with them.

Sometimes retreating to a previous save (like one done at low wiggle power) helps.

Sometimes starting over completely helps. Early folds tend to stick around.

Sometimes tolerating some loss in score helps.

Sometimes lowering the clashing importance helps.

jeff101's picture
User offline. Last seen 23 hours 34 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
More ideas:

Most Recipes (and tools like wiggle & shake) stop gaining points after a while,
but if you try something different for a while and then try the same Recipe or tool again,
you will magically gain points again.

Joined: 06/06/2013
Groups: Gargleblasters
lag and clueless

sporeo and everyone else
I am not a scientist. more of a citizen..... my group can point out obvious faults in my fold. often it is chirality, which is very hard to grasp to us ordinary folk. that said, Jeff's suggestion to back up helps at least some of the time
I set my goals to improve, not to win if that is of any help
Skippy

Joined: 09/24/2012
Groups: Go Science
Desperate tips

In this situation, you probably reached a "local minimum"

This has more chance to occur:
-at High Wiggle Power (turn to high wiggle power as late as possible);
-with local wiggle strategies and recipes (try to use them as late as possible);
-after days of "recipes only" sessions

Suggestions if you still have at least 2-3 days left:

-have a critical look on your solution: is it "beautiful" (aligned sheets with many bonds, orange inside, straight helices, few voids, high filter bonus etc)? if not, give it a hard hand fold in order to correct what you don't like. For example tweaking a helix or a sheet in order to hide the orange sidechains inside of the protein, or moving a structure from one place to another one in order to get better contacts or better rama map or better filter bonus etc.
-then create a new parallel track with this (certainly much low scoring) "B solution" and start again optimizing it with tools and recipes;
-as you'll need a few days in order to have a chance to go over your credited solution, keep trying to rise your credited solution in a parallel track with recipes following suggestions above.

Example for a design puzzle: there are often "Residues are scoring too low" penalty 2-3 days before the end (for example a penalty of -25.61 pts). I simply mutate them (like mutating an aromatic to Histidine) which costs me only 10-20 pts. From this, shake, wiggle all, quickfix then common recipes could help me find a better solution within a day.

Other example for a de novo: I find a disconnected sheet with broken bonds. I rebuild the broken part (usually with a high cost, say -15 000 pts) and I try to fix it again during several days on a parallel track.

Sporeo's picture
User offline. Last seen 1 year 14 weeks ago. Offline
Joined: 01/20/2012
Groups: Go Science
Thanks for all the replies

Thanks for all the replies everyone. I will keep them in mind.

Bruno, what is it about wiggle power that can cause a local minimum? My understanding is that it tightens everything up, but I'm not sure if there's a way to see this has happened in the actual structure. It's been a few years since I've played but I think I've always been tempted to do some of the fine adjustments earlier than I should.

Joined: 10/23/2014
Groups: Contenders
It's kind of hard answering

It's kind of hard answering your questions as there are different types of puzzles which necessitate different approaches.

For example for a puzzle with multiple disulfide bonds your solution may score lower simply because of the bond pairings you chose. For design puzzles some designs will just score better. Lately 2 helices and 3-4 sheets are scoring best. Could be as simple as some players running multiple instances, using faster computers, more computers running 24/7. What recipes to use and when is very important. There are just too many variables. Maybe you can post the puzzle you are having issues with along with explanations and screenshots

I'm also curious on strategies used by various players. Been in the 20's for a while now.

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