AlphaFold Series Advances Peptide Structure Prediction, Highlighting Strengths and Dynamic Limitations

Accurate prediction of peptide structures and peptide-receptor complexes is crucial for rational peptide drug development. However, the inherent conformational flexibility of short and disordered peptides poses a fundamental challenge, as many bioactive peptides adopt their functional conformations only upon binding. This often corresponds to low-probability states that static predictions overlook, leading to inefficiencies in virtual screening and drug design. The AlphaFold model series offers a promising computational platform to address these limitations.

This review synthesizes recent advances across the AlphaFold model series, including AlphaFold2, AlphaFold-Multimer, and AlphaFold3, specifically in the context of peptide structure prediction and applications. The authors systematically discuss the current strengths of these models in predicting monomeric peptide structures and multi-chain complexes, alongside their utility in receptor-binding analysis. Furthermore, the review critically examines the inherent limitations of AlphaFold in capturing dynamic aspects of peptide interactions, such as conformational changes, transient binding events, and the impact of chemical modifications. It also explores integrated computational strategies to enhance predictive accuracy.