search

THUNLP-MT / dyMEAN

This repo contains the codes for our paper "End-to-End Full-Atom Antibody Design"
https://arxiv.org/abs/2302.00203
MIT License
89 stars 8 forks source link

antibody binding inside structure of the target #18

Open keithgmitchell opened 6 months ago

keithgmitchell commented 6 months ago

Hi I love the tool! But i am having some trouble with my antibody and the produced design being inside of the target. Do you have any advice for this?

kxz18 commented 6 months ago

Hi! Thanks for your interest in this work! I think this is a limitation of the current method as I didn't set any features indicating the surface orientation of the epitope. If different random starting states cannot solve the problem, I suggest changing the definition of the epitope (e.g. adding or removing several residues) to induce changing in the outputs. If you are doing research on the ML algorithms, it might also be interesting to explore possible ways to elliminate this problem, such as clash loss, additional surface embeddings and so on.