# Deep generative modeling captures maturation-dependent pairing patterns in human antibodies

## TL;DR

A deep learning model generates antibody light chains from heavy chains, capturing maturation-dependent pairing patterns and improving antibody design.

## Contribution

A two-stage deep learning framework generates plausible antibody pairs from unpaired data, revealing inter-chain dependencies.

## Key findings

- Generated light chains show improved structural quality and germline identity.
- Memory B cell-derived heavy chains produce light chains with restricted V-gene usage.
- Trimodal similarity in generated κ light chains suggests distinct pairing modes.

## Abstract

Understanding antibody heavy-light chain pairing is critical for decoding immune repertoire architecture and designing therapeutic antibodies, yet most sequence databases lack paired chain information. To address this gap, we developed a two-stage deep learning framework. Transformer-based language models were first pre-trained on large corpora of unpaired heavy- and light-chain sequences, then integrated into a sequence-to-sequence model to generate light chains from heavy chain input. Although native light chain recovery was moderate, generated sequences exhibited high germline identity, improved structural quality, and broader framework and complementarity-determining region coverage. Heavy chains from memory B cells generated light chains with more restricted V gene usage, reflecting maturation-dependent selection. Generated κ light chains exhibited a trimodal similarity distribution, indicating distinct functional pairing modes from promiscuous to highly specific. Our approach demonstrates that sequence-to-sequence modeling can uncover inter-chain dependencies and generate plausible antibody pairs, providing a foundation for computational repertoire analysis and therapeutic design.

•A deep generative model enables light chain generation from heavy chains using unpaired data•Conditioning enhances the structural quality and germline coherence of predicted antibodies•Memory B cell-derived heavy chains tend to generate light chains with restricted V-gene use•Generated κ light chains display trimodal similarity, suggesting discrete pairing modes

A deep generative model enables light chain generation from heavy chains using unpaired data

Conditioning enhances the structural quality and germline coherence of predicted antibodies

Memory B cell-derived heavy chains tend to generate light chains with restricted V-gene use

Generated κ light chains display trimodal similarity, suggesting discrete pairing modes

Immunology; Structural biology; Artificial intelligence