Multi-Page PapersVolume 10, Spring 2016

CFHMM: Heterogeneous Tumor CNV Classification by Hidden Markov

Aleksandar Obradovic1,

Hongjian Qi1

1 Department of Computer Science, Columbia University

Abstract

We here develop and implement a Clonal Fraction Hidden Markov Model (CFHMM), to leverage positional information in classifying Tumor CNVs and their corresponding clonal fraction from log-ratio-normalized Tumor/Normal sequencing data. In simulated data, this approach shows accurate calling of CNVs for high-fraction mutations, and improvement in calling over a naïve clustering benchmark across the board, as well as useful purity estimation for dominant clones. 

Availability and Implementation: Source code and documentation is freely available at https://github.com/7lagrange/FCNV implemented in R, with all major operating systems supported.

Contact: azo2104@columbia.edu

Supplementary Information: Additional tables and figures available at https://docs.google.com/document/d/1ohbjWaZ20jXX3Tc64BASuZPWmpnE_ybfwfMjU9jb0mU/edit?usp=sharing