Is Sequence Information All You Need for Bayesian Optimization of Antibodies?

2025-09-29

AI Summary

This study investigates the necessity of structural information and surrogate model selection in antibody Bayesian optimization. Method: We propose the first Bayesian optimization framework integrating structural features with protein language models (PLMs), introducing a novel PLM-driven soft-constraint mechanism that implicitly encodes structural preferences to guide sequence-space search. We systematically compare three modalities—sequence-only, sequence-with-soft-constraints, and sequence-with-explicit-structure—on antibody affinity and stability optimization tasks. Results: Soft constraints enable sequence-only optimization to match the peak performance of structure-dependent methods, lagging only marginally in early optimization stages. Crucially, they substantially reduce reliance on costly 3D structure prediction or experimental determination. Our core contribution challenges the “structural necessity” assumption, establishing PLM-based soft constraints as an efficient, high-fidelity alternative paradigm for antibody design.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Structured Learning
• Constraint Satisfaction and Optimization: Constraint Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Bayesian optimization is a natural candidate for the engineering of antibody therapeutic properties, which is often iterative and expensive. However, finding the optimal choice of surrogate model for optimization over the highly structured antibody space is difficult, and may differ depending on the property being optimized. Moreover, to the best of our knowledge, no prior works have attempted to incorporate structural information into antibody Bayesian optimization. In this work, we explore different approaches to incorporating structural information into Bayesian optimization, and compare them to a variety of sequence-only approaches on two different antibody properties, binding affinity and stability. In addition, we propose the use of a protein language model-based ``soft constraint,'' which helps guide the optimization to promising regions of the space. We find that certain types of structural information improve data efficiency in early optimization rounds for stability, but have equivalent peak performance. Moreover, when incorporating the protein language model soft constraint we find that the data efficiency gap is diminished for affinity and eliminated for stability, resulting in sequence-only methods that match the performance of structure-based methods, raising questions about the necessity of structure in Bayesian optimization for antibodies.

Problem

• Optimizing antibody properties using Bayesian optimization methods
• Incorporating structural versus sequence-only information in optimization
• Evaluating necessity of structural data for antibody engineering

Innovation

• Incorporating structural information into Bayesian optimization
• Using protein language model as soft constraint
• Comparing sequence-only and structure-based optimization methods