predatoR

predatoR predicts the functional impact of missense mutations by transforming protein structures into network models and applying machine learning to classify and prioritize variants of unknown significance (VUS).

Key Features:

• Network-Based Approach: Converts protein structures into network models and computes network properties at mutated sites to capture structural context.
• Machine Learning Integration: Applies machine learning algorithms to network-derived features to predict mutation impact.
• Training and Benchmarking: Trained on VariBench and ClinVar datasets and benchmarked against 32 existing prediction methods using the Missense3D dataset with an AUROC of 0.941.

Scientific Applications:

• VUS classification and prioritization: Provides predictions to classify and prioritize variants of unknown significance for research and clinical review.
• Genomic research: Supports investigation of genotype–phenotype relationships by assessing functional consequences of missense variants.
• Clinical interpretation: Informs variant interpretation to aid diagnostic assessment and support therapeutic decision-making.

Methodology:

Protein structures are transformed into network models and analyzed using machine learning algorithms to evaluate how mutations affect protein function.