ProAffiMuSeq

ProAffiMuSeq predicts changes in binding free energy (ΔΔG) of protein-protein complexes resulting from mutations using sequence-based features and functional classification.

Key Features:

• Sequence-Based Prediction: Predicts changes in binding free energy (ΔΔG) from sequence information without requiring detailed structural data.
• Functional Classification Integration: Incorporates functional class information into the predictive model to improve accuracy and relevance of ΔΔG predictions.
• Performance Metrics: Reports an average correlation coefficient (r) of 0.73 and mean absolute error (MAE) of 0.86 kcal/mol in 10-fold cross-validation, and r = 0.75 with MAE = 0.94 kcal/mol on independent test datasets.
• Comparative Validation: Validated against external datasets, including a blind dataset of 473 mutations (r = 0.27, MAE = 1.06 kcal/mol) and 552 non-redundant interface mutations from SKEMPI 2.0 (MAE = 1.21 kcal/mol), showing performance comparable to structure-based methods.

Scientific Applications:

• Large-scale mutation screening: Enables large-scale analyses of disease-causing mutations in protein-protein interactions when structural data are unavailable or incomplete.
• Disease mechanism inference: Predicts impacts of mutations on binding affinity to infer potential disruptions in cellular functions that could lead to disease.

Methodology:

Integrates sequence-based features with functional classification to model changes in protein-protein interaction affinities (ΔΔG); performance assessed by 10-fold cross-validation and independent testing, including validation against a 473-mutation blind set and SKEMPI 2.0.