Folding Forum

From what I understand from the article, it might be that there's a way to fold more efficiently, thanks to a new 'probability algorithm'.
I mean, it looks to me that it can be applied to Folding, to narrow down hit areas, and focus folding on those, and leave less effective cycles as lower priority.

Our current model still runs too many waste WUs.
If there only was a way to combine this discovery with folding, then folding soon might be a thing of the past.

https://phys.org/news/2019-09-scientist ... roblem.amp

FAH has been using most of that technique for over 10 years. The difference is that step 1 (sample the distribution) is done with AI (which is a useful improvement) rather than with a FAH model. Step 2 (generate a distribution) and step 3 (use the FAH every-atom model starting from interesting points on the statistical distribution) are very much a part of FAH today and have been for 10 years. Search this forum and foldingathome.org for "markov" models.

The only real difference is that FAH was forced to use the every-atom model to perform step 1. Once the points of interest are determined, associated projects are started from those interesting semi-folded states.

See viewtopic.php?f=15&t=31725&hilit=markov and any other discussions regarding "markov" models.