Open denizkavi opened 5 days ago
Also discussed here: https://github.com/sokrypton/ColabFold/issues/76
Hi Deniz,
Certainly it can be achieved in AlphaPulldown. You can use make your selected structures to replace the default template features when running run_multimer_jobs.py
by adding the flags --multimeric_template --description_file=/path/to/descrption.csv
multimeric_template
is a boolean flag that will inform AlphaPulldown to look for customised template structures and replace the original template features. ``´ description_file.csv```should be in the following format:
protein_A,xxx.pdb,C
indicating your query sequence protein_A
should be mapped to chain C your pdb file xxx.pdb
, suppose you already had protein_A.pkl
In addition, the actual sequence of the chain in the PDB does not necessary have to be the same as the protein_A sequence, as AlphaPUlldown will realign the sequences anyway but it will always better if your template covers as many residues as it can.
On the other hand, you can also regenerate the pickle files as described here: https://github.com/KosinskiLab/AlphaPulldown/blob/main/manuals/example_3.md
Yours Dingquan
I have monomer MSAs generated for each sequence in a complex, how might I combine them to be used in a multimer structure prediction? I saw https://github.com/sokrypton/ColabFold/issues/601 discusses this, where the authors propose using monomer structure predictions as custom templates for the multimer structure prediction. But would prefer to maintain the same methodology as AF-Multimer for consistency.