Viability of Extending Approach to Antibodies

[KrajShuffle]

Hi,

This is a great paper and I was wondering whether it is possible to extend this approach of directed evolution to antibodies as well. For instance, I was thinking of using the AbLang2 tokenizer: https://github.com/TobiasHeOl/AbLang2 for tokenizing my antibody sequence and using the masked marginal score as defined for the ESM expert for scoring potential variants. Two things that may be potential issues are that I realized the tokenization is to a one-hot encoded representation and would need to encode the split token (I) as well for distinguishing between the light and heavy chain.

Thanks,
Karthik