David Ryan Koes @dkoes.compstruct.org · Nov 26 Results on LIT-PCBA (an admittedly somewhat problematic benchmark) are quite good, although I personally don't see this replacing more traditional techniques so much as supplementing them. We're looking into using it to initialize "deep docking" runs.
0 likes 2 replies
?
Replies
Have you looked at biases in the model? I ran it on some internal data and it tended to rank the same molecules highly for multiple different bacterial targets. This was a quick-and -dirty test so I’m not sure how to interpret this.
Unlike previous approaches, SPRINT uses attention to learn from embeddings, resulting in some level of interpretability. It turns out the attention mask is sparse with some preference for binding site residues but more for less conserved residues.