DeepMind and Protein Folding

Please confine these topics to things that would be of general interest to those who are interested in FAH which don't fall into any other category.

Moderator: Site Moderators

DeepMind and Protein Folding

Postby Nert » Wed Dec 05, 2018 10:49 pm

Following is a link to an article describing Google's application of their DeepMind platform to protein folding. DeepMind was entered into a competition where participants predict protein structures from a list of their amino acids. Deep Mind came in first.

https://www.theguardian.com/science/2018/dec/02/google-deepminds-ai-program-alphafold-predicts-3d-shapes-of-proteins
Nert
 
Posts: 154
Joined: Wed Mar 26, 2014 7:46 pm

Re: DeepMind and Protein Folding

Postby bruce » Thu Dec 06, 2018 3:32 am

But it cannot predict the unlikely prospect showing how a protein can miss-fold?
bruce
 
Posts: 22249
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: DeepMind and Protein Folding

Postby Nert » Thu Dec 06, 2018 5:32 pm

bruce wrote:But it cannot predict the unlikely prospect showing how a protein can miss-fold?


I think that they are looking for the "normal" results, but that's mostly a guess. I tried to dig through the CASP13 (Critical Assessment of Techniques for Protein Structure Prediction) information linked by the Guardian article, but I couldn't make much headway. The information there is well beyond my superficial understanding of these things. Maybe one of the science guys can provide some high level context.
Nert
 
Posts: 154
Joined: Wed Mar 26, 2014 7:46 pm

Re: DeepMind and Protein Folding

Postby bruce » Thu Dec 06, 2018 5:56 pm

Yes, the CASP is aimed at predicting the normal structure. It's not surprising that AI would be good at that.

FAH goals are different. FAH's goals include a thorough understanding of how and why proteins fold in a specific way and finding the exceptions (known as mis-folding). FAH is also capable of predicting normal structures, but that has never been a primary goal and it's not particularly efficient at doing that.

Many of FAH's recent studies have been broadened to include "normal" interactions, but often exceptions are also discovered which turn out to be very valuable, scientifically.
bruce
 
Posts: 22249
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: DeepMind and Protein Folding

Postby Ricorocks » Fri Jan 04, 2019 5:26 pm

In the early days of computer security, security industry was fragmented, with many small companies. Soon to be gobbled up by larger firms!! Better/worse for that industry can be argued, each having pluses & minuses.

Lot of folding going on, with different approaches to the same common goal. Gobbling up by large companies, would be problematic, does FAH, converse/share/have overlap/could be helpful, with say Rosetta. Or do the two working on the same thing, act proprietary, don't look at my work?

FAH seems to have not participated in CASP, did FAH gleam anything useful from CASP, or just ignored it?
User avatar
Ricorocks
 
Posts: 390
Joined: Thu Aug 04, 2016 1:49 pm
Location: Georgetown, Texas

Re: DeepMind and Protein Folding

Postby bruce » Fri Jan 04, 2019 6:27 pm

FAH and CASP have different goals --- both scientifically useful. They've participated but not aggressively. Finding all possible paths to the most probable end-point is not the most effective way to find that end-point. Or, from tho opposite perspective, finding the most probable end-point is not the most effective way to discover alternate paths which happen to lead to a mis-folded end-point.

Ricorocks wrote:Lot of folding going on, with different approaches to the same common goal. Gobbling up by large companies, would be problematic, does FAH, converse/share/have overlap/could be helpful, with say Rosetta. Or do the two working on the same thing, act proprietary, don't look at my work?


I'm not that familiar with Rosetta so I can't give you a good answer. I have heard Vijay mention University of Washington in a positive way so there's probably some cooperation going on at some level. Like all scientific research, there's always a bit of completion, too -- like we'd like to be able to be able to claim that we made a big discovery first.

(Like the recent PBS special on the discovery of the double-helix in DNA.)
bruce
 
Posts: 22249
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: DeepMind and Protein Folding

Postby Ricorocks » Fri Jan 04, 2019 9:28 pm

Thanks Bruce!
User avatar
Ricorocks
 
Posts: 390
Joined: Thu Aug 04, 2016 1:49 pm
Location: Georgetown, Texas


Return to Scientific discussions (Non-FAH)

Who is online

Users browsing this forum: No registered users and 2 guests

cron