Skip to content

atom_mask will flag real coordinates by chance #4

Description

@FilipSchymik

Hello,

I am using PIPPack for a large scale project and I noticed that a small fraction of my proteins were shorter by one residue after side-chain packing. I checked those backbone pdbs are those seemed to be valid structures. The issue turned out to be the line from inference.py:
atom_mask = (np.sum(atom_positions, axis=-1) != 0.0).astype(np.int32)

I suppose this is a check for placeholder coordinates that are all zero, but in reality some coordinates will sum to zero just by chance, causing the pipeline to drop this position.

You count instead explicitly check for all zero positions, or additionally, consider the "occupancy" column to be 0 as well. Unless of course there is a reason for the sum != 0.0 check.

Thank you,
Filip

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions