Contact prediction for multimeric proteins

[AnjaliSetiya]

Hi all
I'm interested in generating contact maps from the ESM2 attention heads as described in the following github notebook:
https://github.com/facebookresearch/esm/blob/main/examples/contact_prediction.ipynb.
But it is not clear to me if I want to give hetero dimer or multimeric protein as FASTA for predicting contact maps how can I do that. How should the input be given for multi chains for contact map prediction.

[AlejandroSanchezCano]

Hi,

I am facing a similar problem. I want to generate ESM2 embeddings for dimeric proteins. Is appending the sequences a good idea? I have also modelled the 3D structure of these dimers with ESMFold. Could I use those structures to generate ESM2 embeddings?