BenchmarkNCVTools

BenchmarkNCVTools evaluates and compares non-coding variant scoring methods to benchmark their ability to distinguish pathogenic from benign variants in non-coding genomic regions.

Key Features:

• Comparative Analysis: Compares non-coding variant scoring tools including CADD, FATHMM-MKL, Funseq2, GWAVA, DANN, SNP, and SOM scores.
• Benchmarking Against Databases: Benchmarks scoring methods using ClinVar, COSMIC, and the 1000 Genomes Project to assess discrimination between pathogenic and benign variants.
• Performance Insights: Reports that CADD excels in detecting pathogenic variants within protein-coding gene regions using the ClinVar benchmark, while FATHMM-MKL, GWAVA, and SOMliver show superior performance for variants in lincRNAs, pseudogenes, and other non-coding regions when assessed with COSMIC.
• Precision Challenges: Identifies low precision across evaluated methods, indicating the need for additional non-coding genomic features to improve performance.

Scientific Applications:

• Pathogenic variant prioritization: Supports prioritization of pathogenic variants outside coding sequences by comparing deleteriousness scores for non-coding regions.
• Tool selection for non-coding studies: Provides comparative evidence to guide selection of scoring methods for studies of lincRNAs, pseudogenes, other non-coding regions, and protein-coding proximal variants.

Methodology:

Analyses are implemented as Snakemake workflows using C++ and R on Linux systems to ensure efficient and reproducible evaluation.