baker-laboratory / rf_diffusion_all_atom

Public RFDiffusionAA repo
Other
288 stars 44 forks source link

Use sdf/molfile instead of pdb #25

Open VittorioRainaldi opened 3 weeks ago

VittorioRainaldi commented 3 weeks ago

Hi,

is it possible to do diffusion using a small molecule ligand that is not part of a pdb? For example an sdf file downloaded directly from pubchem or even generated in a molecular editor software such as avogadro. I tried converting sdf into pdb but I get an error when the parse_pdb function does not find a sequence in the input pdb.

idx_sm = torch.arange(max(idx_prot),max(idx_prot)+Ls[1])+200 ValueError: max() arg is an empty sequence

As far as I understand the diffusion process is not using the protein sequence at all, so it should be possible to take an isolated ligand as input but I couldn't find a way to do it.

As a side note it would be nice to have some more comments in this code, otherwise it's really hard to figure out what's what.

Best, Vittorio