search

HannesStark / EquiBind

EquiBind: geometric deep learning for fast predictions of the 3D structure in which a small molecule binds to a protein
MIT License
473 stars 109 forks source link

Only ligand sdf file is generated #13

Closed gezmi closed 2 years ago

gezmi commented 2 years ago

Hi,

I might have not understood everything correctly, but I expected the output to be the ligand bound to the receptor. However, I only get the ligand, and it slightly clashes with the receptor and with itself. Is this a normal prediction?

HannesStark commented 2 years ago

Hi, The receptor is just the same as your input receptor and we only treat it as a rigid body.

Unfortunately, it can happen that the predicted ligand has steric clashes with the protein since the model only has a soft constraint in the loss to partially prevent this.

octavian-ganea commented 2 years ago

It is a good idea to fine-tune EquiBind predictions with software like SMINA which might, in some cases, be better at preventing steric clashes. See our EquiBind-S model from the paper which does exactly that.