AgForce Enables Antigen-conditioned Generative Antibody Design

2026-05-20

AI Summary

Existing antibody design methods often neglect antigen structural information when generating complementarity-determining regions (CDRs), leading to antigen-agnostic designs, reduced sequence diversity, and compromised specificity. To address this, this work proposes AgForce, a novel architecture that encodes antigen–antibody complexes using graph neural networks and incorporates framework dropout, gated bottlenecks, and hyperbolic cross-attention mechanisms to suppress antibody-centric shortcuts. The model replaces conventional cross-entropy loss with a hybrid density network combined with annealed multi-choice learning and enforces antigen cycle-consistency constraints to compel reliance on antigen context during CDR generation. Evaluated on CHIMERA-Bench, AgForce achieves state-of-the-art performance in both binding quality and sequence recovery, improving amino acid recovery by 8% over the strongest baseline, outperforming all competitors across interface metrics, and nearly doubling the effective vocabulary size of generated sequences.

Abstract

Antibody design methods condition on antigen structure to generate complementarity-determining regions (CDR), yet a systematic evaluation of baseline methods reveals that they largely ignore the antigen input. We identify three failure modes that explain this behavior. Antigen blindness arises because models derive predictions from antibody framework context rather than antigen information, producing nearly identical CDRs regardless of the target. Vocabulary collapse reduces predicted amino acids to three to five per position, far below the ground truth distribution in native sequences. Moreover, any model trained with standard per-position cross-entropy converges to the positional marginal distribution, making it provably unable to produce antigen-specific sequence predictions. We propose a novel encoder-decoder architecture called AgForce, that uses a graph neural network (GNN) as the encoder and specialized decoders for sequence-structure co-design. Specifically, we apply framework dropout, gated bottlenecks, and hyperbolic cross attention that prevent the antibody shortcut path. In the decoder, a Mixture Density Network (MDN) sequence head with Potts-like pairwise coupling and annealed Multiple Choice Learning (aMCL) replaces the cross-entropy objective with a multi-component distribution whose optimal solution differs from the positional marginal. An antigen cycle consistency head routes gradients through the sequence decoder, forcing predicted distributions to encode antigen identity. AgForce achieves the best binding quality and sequence recovery simultaneously on the CHIMERA-Bench dataset, improving amino acid recovery by 8% over the strongest sequence baseline while surpassing the baselines across all interface metrics, and nearly doubling the effective vocabulary of GNN methods. The source code is available at: https://github.com/mansoor181/ag-force.git

Problem

• antigen-conditioned antibody design
• complementarity-determining regions
• antigen blindness
• vocabulary collapse
• sequence-structure co-design

Innovation

• AgForce
• antigen-conditioned antibody design
• graph neural network
• mixture density network
• cycle consistency