Ferran Gonzalez Hernandez

GenCDR uses LLaMa-based models to generate antibody CDRs with framework context, achieving high recovery and developable designs through alignment-driven training.

This paper introduces a novel corpus and language models for Named Entity Recognition of pharmacokinetic parameters in scientific literature, achieving high accuracy with BioBERT.

The paper benchmarks generative models for antibody sequence design using log-likelihood scores and finds they correlate with binding affinities, suggesting their potential for ranking.