SAAMBE-SEQ

SAAMBE-SEQ predicts changes in protein-protein binding affinity by estimating binding free energy changes (ΔΔG) caused by genetic mutations from protein sequence data.

Key Features:

• Sequence-Based Approach: Operates solely on protein sequence data without requiring structural information.
• Machine Learning Algorithm: Employs a Gradient Boosting Decision Tree model using 80 features that capture evolutionary information, sequence-based characteristics, and changes in physical properties at mutation sites.
• Performance Metrics: Achieved a Pearson correlation coefficient (PCC) of 0.83 in 5-fold cross-validation against experimental ΔΔG and PCC values of 0.37–0.46 in blind tests for complexes lacking structural data.
• Comparative Accuracy: Exhibits accuracy comparable to or exceeding advanced structure-based prediction methods.

Scientific Applications:

• Genomics: Enables high-throughput prediction of mutation effects on protein-protein interactions in genome-scale studies where structures are unavailable.
• Variant Interpretation: Supports interpretation of missense variants by assessing their impact on binding affinities implicated in genetic disorders.
• Therapeutic Prioritization: Aids prioritization of mutations for targeted therapeutic strategies by evaluating their ΔΔG effects on protein-protein binding.

Methodology:

Extracts sequence-based features (80 total) encompassing evolutionary information, sequence characteristics, and physical property changes, and applies a Gradient Boosting Decision Tree model to predict ΔΔG from sequence data.