A highly sensitive and specific workflow for detecting rare copy-number variants from exome sequencing data

Area of Science:

• Genomics
• Bioinformatics
• Medical Genetics

Background:

• Exome sequencing (ES) is a primary diagnostic tool for Mendelian disorders, but detecting germline copy-number variants (CNVs) from ES data is not standard due to performance limitations.
• SNP array is the current standard for genome-wide CNV detection, but ES offers a potentially more comprehensive approach.

Purpose of the Study:

• To comprehensively characterize and improve the performance of an ES-based CNV detection tool, ExomeDepth, for clinical applications.
• To develop and validate a modified ExomeDepth workflow for accurate and reproducible germline CNV detection from ES data.

Main Methods:

• A modified ExomeDepth workflow was developed, excluding low-mappability exons to reduce false positives from repetitive genomic regions.
• An iterative variant calling framework was implemented to assess the reproducibility of CNV detection.
• The modified workflow was evaluated using a cohort of 307 individuals with paired ES and SNP array data, and further tested on targeted STRC gene CNVs in 1972 individuals.

Main Results:

• The modified workflow reduced identified variants by one-third, increasing sensitivity to 97% and improving the false discovery rate to 11.4% compared to the default ExomeDepth pipeline.
• Exclusion of low-mappability exons removed 4.5% of exons, including some in genes difficult for short-read NGS.
• Clinically reported variants showed 100% reproducibility, and targeted STRC gene CNV testing achieved a 100% validation rate.

Conclusions:

• The modified ExomeDepth workflow significantly reduces false positives and enhances the sensitivity and reproducibility of CNV detection from ES data.
• The developed iterative variant calling framework provides a robust method for assessing CNV reproducibility.
• Recommendations are provided for implementing this improved ES-based CNV detection method in clinical settings for Mendelian disorder diagnostics.