search

facebookresearch / esm

Evolutionary Scale Modeling (esm): Pretrained language models for proteins
MIT License
3.16k stars 627 forks source link

Contact-map #26

Closed nasserhashemi closed 3 years ago

nasserhashemi commented 3 years ago

Hi Thanks again for your great repositories; I have a question; how did you calculate contact map in your pre-retrained model? Have you used any supervising datasets or just use the embedded representation of amino acids and pairwise scoring? Thanks Nasser

tomsercu commented 3 years ago

Hi Nasser, Thanks for your question -- It's a logistic regression mapping the attention heads to the contact map, logistic regression weights trained on only 20 proteins. As described in our recent paper to appear at ICLR.

nasserhashemi commented 3 years ago

Thanks so much Great