PON-P: integrated predictor for pathogenicity of missense variants

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput sequencing generates vast amounts of genomic data, necessitating efficient methods for variant effect interpretation.
• Experimental validation methods cannot keep pace with the volume and speed of genomic data generation.
• Computational tools are crucial for assessing the impact of genomic variations on protein function and disease association.

Purpose of the Study:

• To develop and validate an integrated computational pipeline, the Pathogenic-or-Not-Pipeline (PON-P), for predicting the pathogenicity of nonsynonymous genomic variants.
• To leverage and combine the strengths of multiple existing variant prediction tools.
• To provide a reliable method for prioritizing genetic variants for further experimental investigation.

Main Methods:

• Integration of five established variant prediction tools using a random forest methodology.
• Development of the Pathogenic-or-Not-Pipeline (PON-P) for ternary classification of variant effects.
• Performance evaluation through cross-validation and testing on independent datasets, including statistical reliability estimation.

Main Results:

• PON-P demonstrated consistently improved performance in cross-validation and on independent test sets.
• The pipeline provides a ternary classification (pathogenic, benign, uncertain) with a statistical reliability estimate.
• Application to missense variants in a melanoma cell line identified variants in 17 genes predicted to affect protein function, with nine previously linked to cancer pathogenesis.

Conclusions:

• The Pathogenic-or-Not-Pipeline (PON-P) offers a robust and reliable computational approach for predicting the functional impact of nonsynonymous variants.
• PON-P can serve as an effective first-tier screening tool to prioritize genetic variants for downstream experimental validation.
• The pipeline's findings in melanoma cell lines highlight its potential utility in cancer genomics research for identifying disease-associated variants.