A Bayesian Feature Allocation Model for Tumor Heterogeneity

• Activity Number: 169
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Juhee Lee*+ and Peter Mueller and Yuan Ji and Kamalakar Gulukota
• Companies: University of California, Santa Cruz and University of Texas at Austin and NorthShore University HealthSystem and NorthShore University HealthSystem
• Keywords: Cell types ; Feature allocation models ; Indian buffet process ; Markov chain Monte Carlo ; Next-generation sequencing ; Random binary matrices

Abstract

We propose a feature allocation model to model tumor heterogeneity. The data are next-generation sequencing data (NGS) from tumor samples. We use a variation of the Indian buffet process to characterize latent hypothetical subclones based on single nucleotide variations (SNVs). We define latent subclones by the presence of some subset of the recorded SNVs. Assuming that each sample is composed of some sample-specific proportions of these subclones we can then fit the observed proportions of SNVs for each sample. By taking a Bayesian perspective the proposed method provides a full description of all possible solutions as a coherent posterior probability model for all relevant unknown quantities including the binary indicators that characterize the latent subclones by selecting (or not) the recorded SNVs, instead of reporting a single solution.