msa to use in alphafold

Question

[andreapamela.aguilar]

Hello

I am learning to use alphafold. I am using the colab version. I want to make the prediction of an antigen-antibody complex. I have been doing it but the antibody binds to any side and not where it should (I have that experimental information). In several articles I read that this problem can be solved using evolutionary information by loading an msa. I put together the sequences of complexes to use (that have their pdb) that bind to the region. The sequences have the required identity. However, in this case it is not detailed how to assemble this msa of antigen-antibody complexes. And I still have problems in this step. Could someone tell me how to do it?

[Pierre Lindenbaum]

please review the questions you asked before and comment/validate the answers: CRISPRCasFinder and Anaconda ; Is there an online server to get a great phylogenetic tree? ; how to create a consensus sequence from different assemblies

[dthorbur]

I'm only just venturing into this kind of work myself, so I'm hoping someone with more knowledge comes along soon.

I would suggest looking into the the OpenFold project and methods in papers like this that discuss some of the considerations for generating MSAs for AF2.

You can assemble an MSA in A3M format quite easily which can be used by colabfold_batch in local colabfold (unsure about notebook versions). mmseqs and hh-suite can generate appropriate MSAs, and to my knowedge the PDB template directory should be easy to add to if you add the --template flag to the batch function.

What I am unsure of is if there is a requirement for using certain databases. OpenFold requires one of the database MSAs by default, and I'm unsure if colabfold does too.