COPCNVBD: An Integrated Approach for Somatic Copy Number Variation and Breakpoint Detection Using Whole Genome Sequencing Data

Read depth (RD) signals anomaly-based copy number variation (CNV) detection methods using whole genome sequencing data are affected by the measurement scale and parameters, and the breakpoint of CNVs is easily influenced by the sizes of windows in preparing RD signals. In this study, we propose an integrated approach for somatic CNV and breakpoint detection, COPCNVBD, which builds an improved Copula-Based Outlier Detector (COPOD) anomaly detection method to infer approximate locations of CNVs without hyper-parameters. Then, COPCNVBD first regards the precise CNV breakpoint location as the image boundary detection and makes use of the pair-end mapping (PEM) reads information to design a CNV breakpoint identification strategy. We use simulation datasets with different tumor purity and coverage settings and real samples to demonstrate the performance of COPCNVBD and compare it with peer-popular tools. Simulation results show that COPCNVBD has a superior comprehensive performance even on low coverage data with low tumor purity. Results on three real cancer samples show that our proposed COPCNVBD is able to detect moderate CNVs and has a high consistency. The proposed COPCNVBD can be as a tool for the analysis of CNVs in the genome.