format of the msas for antibody/antigen pairs

[avilella]

Hi,

I'd like to supply the boltz prediction command with the msas for the antibody (fv pair with a linker) and antigen (single chain) msas so it speeds up the calculation rather than using the msa server.

I noticed that there are several msas produced as output when running the msa server, some under a folder called 'paired' and others under 'unpaired'.

It seems like for a query with protein A and protein B, it produces 'unpaired' a3m files where 101 and all the hits go first, then there is a ^@ character, and then the 102 and all the hits go afterwards, and the file also ends with ^@.

How do I compose my own a3m files in cases like this? Does a single unpaired a3m file suffice? Does the parser care about the ^@ characters?

Thanks

[RoKant]

Very interested in this as well, trying to run many similar antibody sequences against the same protein. Run into the rate limits of the MMseqs2 server and speed issues. Can refer to .a3m files for the separate components, but the paired msa will be important I suppose.