SAMPDI-3D

SAMPDI-3D predicts changes in protein-DNA binding free energy resulting from mutations to quantify effects on protein-DNA interactions for interpretation of mutation impacts and molecular disease mechanisms.

Key Features:

• Machine learning algorithm: Uses a gradient boosting decision tree to predict changes in binding free energy.
• Comprehensive mutation analysis: Considers mutations in both the DNA-binding protein and the DNA sequence.
• Physicochemical and structural features: Incorporates physicochemical properties, structural aspects of mutation sites, and protein-DNA interaction dynamics as input features.
• Predictive performance: Achieves Pearson correlation coefficients of 0.76 for protein mutations and 0.80 for DNA mutations against experimentally determined binding free energy changes.
• Benchmark superiority: Outperformed existing state-of-the-art methods in blind benchmark tests using three datasets collected from the literature.
• Speed and scalability: Implemented for rapid analysis enabling genome-scale investigations.

Scientific Applications:

• Molecular disease mechanism elucidation: Quantifies mutation impacts on protein-DNA binding free energy to support identification of molecular origins of disease.
• Drug development: Provides quantitative estimates of how mutations modulate protein-DNA interactions useful for designing modulatory compounds.
• Genomic studies: Enables large-scale analysis of mutation effects across genomes.

Methodology:

Gradient boosting decision tree models trained on features including physicochemical properties, structural features of mutation sites, and protein-DNA interaction dynamics were evaluated against experimentally determined binding free energy changes in blind benchmark tests on three literature datasets, yielding Pearson correlations of 0.76 (protein mutations) and 0.80 (DNA mutations).