Discrimination of recurrent CNVs from individual ones from multisample aCGH by jointly constrained minimization

Copy number variations (CNVs) are associated with complex diseases and particular tumor types; thus, reliable CNV identification has substantial potential value. Several different high-throughput technologies can be used to identify CNV sites. One commonly used approach to detect CNVs is array-based comparative genomic hybridization (aCGH). Recent advances in sequencing technology make it affordable to obtain aCGH data for multiple samples, and an increasing number of methods have become available for detecting recurrent CNV regions across samples. However, copy number is highly dynamic in cancer cells and thus individually specific. In contrast, researchers anticipate that detecting recurrent CNVs in samples is an indication that the tumors share the same origin and thus possibly also have common oncogene drivers and tumor insurgence. Therefore, accurate discrimination of recurrent from individual CNVs is vital to explain various phenotype differences and genetic diseases. To address this problem, we present a general model to identify and discriminate two types of CNVs, namely sample-wised individual and group-wised recurrent CNVs , from multi-sample aCGH profiles. We first imposed general assumptions on the sample-wised and group-wised CNVs. Then, we detected CNVs by proposing a convex optimization with multi-constraints to distinguish the two CNV types. An efficient numerical algorithm was then presented to solve the problem. We demonstrated the performance of the proposed method by comparing the results with those of popular alternative methods on both simulated and empirical breast cancer datasets. The experimental results demonstrate that the proposed method outperformed its peers by successfully detecting CNV patterns and accurately discriminating their differences.