IgPose: A Generative Data-Augmented Pipeline for Robust Immunoglobulin-Antigen Binding Prediction

2026-03-16

AI Summary

This study addresses the challenge of immunoglobulin–antigen (Ig–Ag) binding prediction, which is hindered by the scarcity of experimental structures and limited accuracy in antibody modeling. To overcome these limitations, the authors propose IgPose, a generative data augmentation framework that enriches training data with high-fidelity decoy conformations and integrates geometric and evolutionary features for binding pose identification and scoring. Key innovations include the first Structure-based Immunoglobulin Decoy Database (SIDD), an interface-focused k-hop sampling strategy combined with biologically guided pooling, and a twin-network architecture leveraging equivariant graph neural networks, ESM-2 embeddings, and gated recurrent units for conformation classification and DockQ scoring, respectively. IgPose significantly outperforms existing physics-based and deep learning methods on both an internal test set and the CASP-16 benchmark, offering an efficient and accurate tool for high-throughput antibody discovery.

Technology Category

• Computer Vision: Biometrics, Face, Gesture & Pose
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Predicting immunoglobulin-antigen (Ig-Ag) binding remains a significant challenge due to the paucity of experimentally-resolved complexes and the limited accuracy of de novo Ig structure prediction. We introduce IgPose, a generalizable framework for Ig-Ag pose identification and scoring, built on a generative data-augmentation pipeline. To mitigate data scarcity, we constructed the Structural Immunoglobulin Decoy Database (SIDD), a comprehensive repository of high-fidelity synthetic decoys. IgPose integrates equivariant graph neural networks, ESM-2 embeddings, and gated recurrent units to synergistically capture both geometric and evolutionary features. We implemented interface-focused k-hop sampling with biologically guided pooling to enhance generalization across diverse interfaces. The framework comprises two sub-networks--IgPoseClassifier for binding pose discrimination and IgPoseScore for DockQ score estimation--and achieves robust performance on curated internal test sets and the CASP-16 benchmark compared to physics and deep learning baselines. IgPose serves as a versatile computational tool for high-throughput antibody discovery pipelines by providing accurate pose filtering and ranking. IgPose is available on GitHub (https://github.com/arontier/igpose).

Problem

• immunoglobulin-antigen binding
• data scarcity
• structure prediction
• binding pose identification
• computational antibody discovery

Innovation

• generative data augmentation
• equivariant graph neural networks
• immunoglobulin-antigen binding
• interface-focused sampling
• synthetic decoy database