MPFIL: Multi-Level Protein Feature Integration for Predicting Mutation-Induced Protein-Protein Binding Affinity Changes

MPFIL predicts mutation-induced changes in protein–protein binding affinity (ΔΔG) by integrating structural and sequential features, achieving competitive and consistent performance across three benchmark datasets.

• Why it matters: Mutation effects can alter both local residue environments and the structural organization of protein complexes, and prior handcrafted feature methods add cost and domain-specific requirements.
• What they did: MPFIL uses structural feature extraction to capture mutation-induced structural changes and wild-type versus mutant differential information, plus sequence modules for local context and long-range dependencies, then integrates features for ΔΔG prediction.
• The result: Across three benchmark datasets, MPFIL shows competitive, consistent results, and ablation studies confirm structural and sequential features each contribute to modeling mutation-induced ΔΔG changes.