Some ideas for enhancements...

[ebettler]

Hi,
nice work. I'll test it soon. Before I do that, though, I have some questions and/or suggestions for improvements:

• Why are you using SIGNALP5 instead of the newest version, SIGNALP6? SIGNALP6 can predict the presence of signal peptides and the location of their cleavage sites in proteins from Archaea, Gram-positive Bacteria, Gram-negative Bacteria, and Eukarya.
• You could use MMSEQ to cluster proteins and reduce the number of sequences to manage.
• You could test if SignalP and propept are already known using Uniprot numbers instead of running SignalP (I wrote a script for that).
• Can we run PPIFold post-analyse on AlphaPullDown data already available ?

[Qrouger]

Hi, Thanks for all these ideas.
For now the main reason we use SignalP5 it's due to our study model. We work on Gram- organisms and sometimes fast SignalP6 failsto detect signal peptide (ex : Q6I6C7). But, we plan to incorporate SignalP6 to offer users a choice.

MMSEQ is used in the feature generation with the new version of Alphapulldown.

Ok, that's a good idea. You use a database ?

Cuurently, the only way to perform post-analys with pre-generated data is to place all your model in the right directory (result_all_vs_all, result_homo_oligo) as show in the example. Then run PPIFold as if models are not generated, and it automatically skip model generation (or will generate missing models).
A problems is the version of Alphapulldown you use to generate your models.

Quentin