Computational design of protein complexes: influence of binding affinity

The interaction of proteins with diverse molecular partners, including other proteins, nucleic acids, and carbohydrates, is essential for performing various functions, from signal transduction and gene regulation to immune recognition and cellular transport. These interactions are largely governed by the three-dimensional structures and dynamics of biomolecular complexes, which in turn dictate their binding affinities and functional specificity. While recent advances in AI-driven structure prediction have greatly improved our ability to model such complexes, accurately predicting and engineering their binding affinities remains a key challenge. In this article, we review emerging computational strategies for affinity prediction and rational design across protein-protein, protein-DNA/RNA, and protein-carbohydrate complexes. We discuss the role of machine learning and deep learning in advancing structure-based and sequence-based affinity models, assess current databases and benchmarks, and highlight recent tools for predicting the effects of mutations on binding affinity. We conclude by discussing future opportunities at the intersection of AI, high-throughput screening, and data-driven modeling to enable affinity-guided design of functional biomolecular assemblies.