Summary
Explicitly separating germline from non-germline residues improves zero-shot antibody-affinity prediction. PRISM likelihoods correlated positively with experimental affinity across three deep-mutational-scanning landscapes on which the compared antibody language models were anticorrelated (1).
See also
- Explicit germline encoding improves prediction of non-germline CDRH3 residues
- Generic PLMs outperform antibody-specific PLMs on zero-shot predictions of affinity changes
1.
Kim J, Blalock N, Kulkarni A, Nakamura K, Romero PA. Explicit representation of germline and non-germline residues improves antibody language modeling. openRxiv; 2026. Available from: https://doi.org/10.64898/2026.05.06.723387