# Paraplume: A fast and accurate antibody paratope prediction method provides insights into repertoire-scale binding dynamics

## TL;DR

Paraplume is a fast method to identify antibody binding regions from sequences, enabling large-scale analysis of immune responses and antibody evolution.

## Contribution

Introduces Paraplume, a sequence-based paratope prediction method using protein language models that outperforms existing methods and scales to large repertoires.

## Key findings

- Paraplume achieves superior performance in paratope prediction compared to current methods without requiring structural input.
- Antigen-specific mutations correlate with larger paratopes, suggesting a mechanism for affinity enhancement.
- Paraplume improves downstream tasks like binder classification and epitope binning by reweighting protein language model embeddings.

## Abstract

The specific region of an antibody responsible for binding to an antigen, known as the paratope, is essential for immune recognition. Accurate identification of this small yet critical region can accelerate the development of therapeutic antibodies. Determining paratope locations typically relies on modeling the antibody structure, which is computationally intensive and difficult to scale across large antibody repertoires. We introduce Paraplume, a sequence-based paratope prediction method that leverages embeddings from protein language models (PLMs), without the need for structural input and achieves superior performance across multiple benchmarks compared to current methods. In addition, reweighting PLM embeddings using Paraplume predictions yields more informative sequence representations, improving downstream tasks such as binder classification and epitope binning. Applied to large antibody repertoires, Paraplume reveals that antigen-specific somatic hypermutations are associated with larger paratopes, suggesting a potential mechanism for affinity enhancement. Our findings position PLM-based paratope prediction as a powerful, scalable alternative to structure-dependent approaches, opening new avenues for understanding antibody evolution.

Accurately identifying the small region of an antibody that binds the target antigen, the paratope, is important for immune recognition and designing effective therapies. Most existing approaches depend on 3D structural modeling, which is computationally demanding and limits large-scale analyses. We present a fast and scalable method that predicts paratopes directly from antibody sequences using protein language models. We show that asymmetric paratopes reflect biological binding mechanisms and correlate with the structures of cognate antigen epitopes. Applying our method to antibody repertoires, we find that affinity maturation in response to antigen exposure is associated with an increase in predicted paratope size. Our results open up new directions in exploring the functional consequences of antibody diversification and evolution.