A Hidden Markov Model for Analyzing Allele-Specific Copy Number Alteration Accounting for Hypersegmentation in Next Generation Sequencing

• Activity Number: 68 - Government Health Statistics
• Type: Contributed
• Date/Time: Sunday, July 30, 2017 : 4:00 PM to 5:50 PM
• Sponsor: Government Statistics Section
• Author(s): Hyoyoung Choo-Wosoba* and Paul S. Albert and Bin Zhu
• Companies: NIH/NCI-DCEG-Biostatistics Branch and National Cancer Institute/NIH and NIH/NCI-DCEG-Biostatistics Branch
• Keywords: hidden Markov model ; cancer genomics ; Next Generation Sequencing ; E-M algorithm ; allele-specific copy number alteration

Abstract

Somatic copy number alternation (SCNA) is a common feature of cancer genome and is associated with cancer etiology and prognosis. The allele-specific SCNA analysis aims to identify the allele-specific copy numbers over the entire chromosomes as well as the ploidy and the tumor purity for a tumor sample. Next high resolution generation sequencing platforms produce abundant read counts across the exome or whole genome that are potentially susceptible to hypersegmentation, a phenomenon where numerous regions with very short length are falsely identified a SCNA. In this regard, we propose a hidden Markov model approach that accounts for hyper-segmentation for allele-specific SCNA analysis. We propose an efficient E-M algorithm procedure that uses a forward-backward algorithm for evaluating the E-step. We demonstrate the robustness of our method in simulations and with a renal cell carcinoma sample from the Cancer Genome Atlas (TCGA) study. We further extend the model by considering subclones in tumor samples, which provides the insight into the tumor evolutions underlying intra-tumor heterogeneity.