Conformation-Aware Structure Prediction of Antigen-Recognizing Immune Proteins

2025-07-11

AI Summary

This study addresses key bottlenecks in conformational prediction of antigen-recognition immune proteins—namely, antibodies, nanobodies, and T-cell receptors—including inaccurate modeling of bound (holo) versus unbound (apo) states and weak variable-domain reconstruction. We propose the first lightweight deep learning model explicitly distinguishing apo and holo states. Methodologically, leveraging a high-resolution private structural dataset, we employ a tokenization-based training strategy specifically optimized for variable-domain conformational modeling; critically, we embed state priors directly into the network architecture, substantially improving out-of-distribution generalization. Experiments demonstrate state-of-the-art accuracy across diverse immune protein classes, with computational cost reduced by an order of magnitude relative to general-purpose tools such as AlphaFold2. The framework establishes a new paradigm for efficient, high-precision conformational prediction in macromolecular drug design.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Large Vision Models
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

We introduce Ibex, a pan-immunoglobulin structure prediction model that achieves state-of-the-art accuracy in modeling the variable domains of antibodies, nanobodies, and T-cell receptors. Unlike previous approaches, Ibex explicitly distinguishes between bound and unbound protein conformations by training on labeled apo and holo structural pairs, enabling accurate prediction of both states at inference time. Using a comprehensive private dataset of high-resolution antibody structures, we demonstrate superior out-of-distribution performance compared to existing specialized and general protein structure prediction tools. Ibex combines the accuracy of cutting-edge models with significantly reduced computational requirements, providing a robust foundation for accelerating large molecule design and therapeutic development.

Problem

• Predicts structures of antigen-recognizing immune proteins accurately
• Distinguishes bound and unbound conformations for precise modeling
• Enables efficient large molecule design with reduced computation

Innovation

• Ibex predicts bound and unbound immune protein conformations
• Trains on labeled apo and holo structural pairs
• Combines high accuracy with low computational cost