sokrypton / ColabDesign

Making Protein Design accessible to all via Google Colab!
529 stars 118 forks source link

MPNN designed sequence not separated correctly with multiple targets and multiple binders #141

Open data2code opened 1 year ago

data2code commented 1 year ago

With a toy example, my contig string is 'A 20 E 10' with a modified pdb structure. The output sequence is: TTCCPSIRSNFNVCRLPGTPEAICATYTGCIIIPGATCPGDYAN/RRRNNDKPVDMLYPMVAMEMTLGSEFEVME

In the above output, the two target sequences (chain A and E) are combined, followed by "/", then followed by two binder sequences (20 + 10). There is no separation "/" between the two targets and the two binders.

Some initial tracing of the code points to line 27 in af.prep._prep_binder

self._lengths = [self._target_len, self._binder_len]

self._lengths only contain two elements. This is later passed on to the mpnn code, leading to mpnn think there are only two chains in sequence output. I can certainly read the sequences from the final PDB file, just wonder if this is something that needs to be fixed.

Thanks!

sokrypton commented 1 year ago

This shouldn't affect the results. but I'll make some edits so that the chains are printed as separate chains.

Thanks!

On Sat, May 13, 2023, 1:44 AM Yingyao Zhou @.***> wrote:

With a toy example, my contig string is 'A 20 E 10' with a modified pdb structure. The output sequence is: TTCCPSIRSNFNVCRLPGTPEAICATYTGCIIIPGATCPGDYAN/RRRNNDKPVDMLYPMVAMEMTLGSEFEVME

In the above output, the two target sequences (chain A and E) are combined, followed by "/", then followed by two binder sequences (20 + 10). There is no separation "/" between the two targets and the two binders.

Some initial tracing of the code points to line 27 in af.prep._prep_binder

self._lengths = [self._target_len, self._binder_len]

self._lengths only contain two elements. This is later passed on to the mpnn code, leading to mpnn think there are only two chains in sequence output. I can certainly read the sequences from the final PDB file, just wonder if this is something that needs to be fixed.

Thanks!

— Reply to this email directly, view it on GitHub https://github.com/sokrypton/ColabDesign/issues/141, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA76LAUH5ZU2YGMEATORZTLXF4NUZANCNFSM6AAAAAAYAJCKPQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

data2code commented 1 year ago

Yes, the results are correct. Thanks!