|
|
|
|
|
|
by D-Machine
192 days ago
|
|
|
Yeah this article is garbage. The real problem with protein-folding is not compute, or training on known configurations only, but figuring out a differentiable loss that is related to the energy configuration of generated new sequences / molecules, and iterative folding and all sorts of other things. It is very much NOT just a "throw lots of data at GPUs" problem. This is all covered cursorily even by Wikipedia - https://en.wikipedia.org/wiki/AlphaFold#AlphaFold_2_(2020). |
|
|