A Sparse Model Based Detection of Copy Number Variations From Exome Sequencing Data

Abstract

Goal Whole-exome sequencing provides a more cost-effective way than whole-genome sequencing for detecting genetic variants such as copy number variations (CNVs). Although a number of approaches have been proposed to detect CNVs from whole-genome sequencing, a direct adoption of these approaches to whole-exome sequencing will often fail because exons are separately located along a genome. Therefore, an appropriate method is needed to target the specific features of exome sequencing data. Methods In this paper a novel sparse model based method is proposed to discover CNVs from multiple exome sequencing data. First, exome sequencing data are represented with a penalized matrix approximation, and technical variability and random sequencing errors are assumed to follow a generalized Gaussian distribution. Second, an iteratively re-weighted least squares algorithm is used to estimate the solution. Results The method is tested and validated on both synthetic and real data, and compared with other approaches including CoNIFER, XHMM and cn.MOPS. The test demonstrates that the proposed method outperform other approaches. Conclusion The proposed sparse model can detect CNVs from exome sequencing data with high power and precision. Significance Sparse model can target the specific features of exome sequencing data. The software codes are freely available at http://www.tulane.edu/wyp/software/ExonCNV.m

Preprocessing Sequence Coverage Data for More Precise Detection of Copy Number Variations

“…Also, hybridization in WES causes low or no coverage in some regions of the genome, which introduces more mappability bias. These issues need to be considered for an appropriate CNV detection method for WES data [7].…”