Question about how to interpret MaSIF-site results as residue-level hotspots

[seojinnny]

Hello, and thank you for making MaSIF available.

I understand that the issue tracker is mainly intended for reporting problems,
so I feel a bit cautious about asking a question here, but I would like to ask for your guidance on the following.

I am currently using MaSIF-site to identify hotspot candidates on a protein surface,
and I would like to interpret them at the residue level so that they can be used for downstream binder design tasks.

I understand that the .npy files from MaSIF-site represent scores for surface vertices, and that they correspond to the .ply files. However, I have a few questions about how to connect these vertex-level scores to residues in the original PDB structure and interpret them as residue-level hotspots.

1. Would it be an appropriate approach to use the .npy scores, the .ply mesh, and the original PDB structure together,
   and map each vertex to the nearest residue?

2. When calculating residue-level scores, would it be more appropriate to use the maximum of the vertex scores mapped to each residue, or is it more common to use the mean?

3. When selecting hotspot residues, is there a recommended cutoff for MaSIF-site scores, or is it more common to select the top N residues rather than using an absolute threshold?

4. When interpreting hotspots at the residue level, I would also like to ask whether there is a more recommended approach than simply mapping vertices to residues using the .npy and .ply outputs. In addition, if there are any other output files or sources of information beyond .npy and .ply that are useful for residue-level interpretation, I would greatly appreciate your advice.

Any related experience or recommendations would be a great help.

Thank you for reading.