EMMVEP: An Ensemble Method for Protein Missense Variant Effect Prediction Based on Multi-Source Feature Fusion.

Area of Science:

• Genomics
• Computational Biology
• Protein Science

Background:

• Missense mutations are common genetic variations that can alter protein function, posing challenges in distinguishing pathogenic from benign variants.
• Accurate prediction of missense mutation effects is crucial for understanding genetic diseases and guiding clinical decisions.

Purpose of the Study:

• To introduce EMMVEP, an ensemble-based computational approach for predicting the functional impact of protein missense mutations.
• To evaluate EMMVEP's performance against existing variant effect prediction methods.

Main Methods:

• EMMVEP integrates diverse features including protein sequence information, physicochemical properties from AlphaFold, and allele frequencies from gnomAD.
• Categorical boosting is employed to build the ensemble model for predicting mutation pathogenicity.

Main Results:

• EMMVEP achieved high performance on a benchmark dataset, with an Area Under the Curve (AUC) of 0.907 and an Area Under the Precision-Recall Curve (AUPR) of 0.879.
• The method outperformed 20 general variant effect prediction tools.
• Pathogenicity probabilities for over 216 million potential amino acid substitutions across 19,233 human genes are provided.

Conclusions:

• EMMVEP offers a robust and accurate method for predicting missense mutation effects, enhancing the interpretation of genetic variants.
• The tool provides valuable insights for identifying pathogenic mutations, with significant implications for both research and clinical applications.