Accurate identification of significant aberrations in contaminated cancer genome

Somatic Copy Number Alterations (CNAs) are quite common in human cancers. Identifying CNAs and Significant Copy number Aberrations (SCAs) in cancer genomes is a critical task in searching for cancer-associated genes. The advanced genomic technologies, such as SNP array technology, facilitate copy number study at a genome-wide scale with high resolution. However, in reality, due to normal tissue contamination, the observed intensity signals are actually the mixture of copy number signals contributed from both tumor cells and normal cells. This genetic heterogeneity could significantly affect the subsequent copy number analysis and SCAs detection. In order to accurately identify significant aberrations in contaminated cancer genome, we devise an approach including two major steps. We first use a statistical method, Bayesian Analysis of Copy number Mixtures (BACOM) to estimate the normal tissue contamination fraction and recover the "true" copy number profile. Then, based on the recovered profiles, we detect SCAs using Genome-wide Identification of Significant Aberrations in Cancer Genome (SAIC). We comprehensively evaluate the performance of the proposed algorithm on a large number of simulation data. The results show that the algorithm has higher detection power than peer methods including the most popular GISTIC. We then apply the method to the real copy number data of Glioblastoma Multiforme and successfully identified majority of SCAs reported by GISTIC, and some novel SCAs that contain some cancer-associated genes.