Prediction of Tumor Subtypes by Mixture Modeling of Somatic Mutation Profiles

• Activity Number: 84
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Lin Hou*+ and Hongyu Zhao
• Companies: Yale School of Public Health and Yale University
• Keywords: mixture model ; sub-clone ; cancer genomics ; computational biology ; statistical genomics ; next generation sequencing

Abstract

Identifying tumor subtypes is important for diagnosis and therapy in tumor treatment. Microarray-based mRNA expression profiling is commonly used to classify tumors into subtypes. One limitation of this approach is that, while the sub-clone issue is frequently observed and important with clinical implications, it is difficult to separate clonal and sub-clonal signals based on microarray data. With the development of sequencing technology, many tumor types have been sequenced with sufficient depth to identify both clonal and sub-clonal somatic mutations in individual patients, including ovarian cancer, breast cancer, glioblastoma, etc., revealing a high-resolution mutational landscape of tumor cells. Experimental data showed that there are several mutation signatures in breast cancers, reflecting the mutational process underlying tumorigenesis. In this presentation, we will first motivate the biological problem with cancer sequencing data, describe a mixture model for clonal and sub-clonal somatic mutation profiles (with information of allele frequency, contamination rate of normal cells, and copy number variation), and demonstrate the performance of our method to real data.