Evaluation of somatic copy number estimation tools for whole-exome sequencing data

Area of Science:

• Genomics
• Cancer Research
• Bioinformatics

Background:

• Whole-exome sequencing (WES) is a standard for genetic variant detection in diseases.
• WES data can identify copy number variations (CNVs) but face challenges due to uneven read coverage.
• Developing robust WES-based CNV detection tools remains an active research area.

Purpose of the Study:

• To evaluate the performance of six WES somatic CNV detection tools.
• To compare WES-based CNV calls against a reference set derived from SNP array and whole-genome sequencing.
• To identify limitations and areas for improvement in current WES CNV analysis algorithms.

Main Methods:

• Evaluation of six WES somatic CNV detection tools: ADTEx, CONTRA, Control-FREEC, EXCAVATOR, ExomeCNV, and Varscan2.
• Utilized WES data from 150 cancer patients (kidney chromophobe, bladder urothelial carcinoma, stomach adenocarcinoma) from The Cancer Genome Atlas.
• Compared WES CNV calls against a reference CNV set generated by SNP array 6.0 and whole-genome sequencing.

Main Results:

• Significant variability observed among the CNV calls from the six WES tools and the reference set.
• Overlap between WES CNV calls and the reference set ranged from 13-77% based on a 50% overlap criterion.
• High rates of discordant calls, including gains called as losses (vice versa), and substantial differences in CNV size and number were noted.
• ADTEx and EXCAVATOR demonstrated the best performance in terms of precision and sensitivity.

Conclusions:

• Current algorithms for somatic CNV detection from WES data exhibit performance limitations and high variability.
• Significant discrepancies exist between WES-based CNV detection and reference methods, impacting accuracy.
• There is a clear need for the development of more robust and reliable algorithms for WES-based somatic CNV analysis.