Good ideas and conversation. No ads, no tracking. Login or Take a Tour!
I'm curious if there are ways of building folding programs that accelerate MD approaches using LTSM/RNN type neural network models. A literature review hasn't turned up anything related, although maybe in this case people have tried and failed. The closest thing I found was this paper on employing constraints in the MD simulations. The idea would be to train models that behave like human players of foldit. A friend of mine regards this as extremely uninteresting because it seems to him to be a case of incremental improvement. He hasn't been frustrated repeatedly by our lack of capability in this space. It would be a dream to do constructive biology on computers.