Shape similarity

[drewnutt]

Is the shape similarity computed with shape_screen included in the original manuscript? All I can find in the paper for the pairwise scoring is the use of three interaction types for protein-ligand interaction similarity (hydrogen bonds, salt bridges, and hydrophobic) and the maximum common substructure RMSD between pairs of poses.

If this shape similarity was not included in the original manuscript, do you have any docking results comparing using the shape similarity vs not using the shape similarity in the ComBind score?

[jpaggi]

Shape similarity was not included in the pose prediction benchmarks in our paper.

We did however include it in the virtual screening benchmarks. For virtual screening, it was helpful as a replacement to the substructure similarity metric, which isn't really applicable for ranking diverse molecules because it can only be applied in cases where a shared substructure exists.

I have not gone back to check whether adding shape similarity improves pose prediction performance.