This is cool. Although, there's not enough emphasized how the training is done. Particularly, what exactly AF2 learns from all that PDB structures that it consumes during the training stage? And all that 170M parameters in the system, what they are? Do they contain some representation of protein packing principles or what?
It is extremely nice illustration of protein structure. This lecuture provides us a good method for predicting protein struture in a wise manner. It provides a good model like alpha fold. Customer, most of us are familiar with Watsons -Crick helical folding Model. Compared these traditional models, this methods gives enables Highly accurate protein structure prediction with AlphaFold. Thanks!
This is ridiculous! God bless your hearts. Thank you for your dedication to science and education 🙏 The only constant is change
It is, yeah.
the explainability efforts, in which you described 48 predictors at each layer, for each moment of training, love this idea
This is cool. Although, there's not enough emphasized how the training is done. Particularly, what exactly AF2 learns from all that PDB structures that it consumes during the training stage? And all that 170M parameters in the system, what they are? Do they contain some representation of protein packing principles or what?
It is extremely nice illustration of protein structure. This lecuture provides us a good method for predicting protein struture in a wise manner. It provides a good model like alpha fold. Customer, most of us are familiar with Watsons -Crick helical folding Model. Compared these traditional models, this methods gives enables Highly accurate protein structure prediction with AlphaFold. Thanks!
I am really looking forward to the next version of alpha fold. Thanks for great explanation!
52:20 There is literally an example, where some guy got 1 chain out of 40 in pdb structure backwards, LOL! It was fixed using Alphafold.
can it re-configure a world?
amzng. Perhaps the next project shoud be aging. Just a suggestion