A Robust Statistical Framework to Whole-Genome Outlier Identification for Characterizing Structural Variants

• Activity Number: 340
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Kylie Ainslie*+ and Jeanne Kowalski
• Companies: and Emory University
• Keywords: gene expression outliers ; change point model ; next generation sequencing

Abstract

Cancer is characterized by genetic instability, such as translocations. Due to the heterogeneous nature of cancers, traditional analytical methods to determine genes exhibiting translocation events fail. Recent next generation sequencing (NGS) data has made it clear that in the average cancer genome hundreds of genes are affected by structural changes via translocations, among other mechanisms. The challenge is to identify driving gene disruptions amongst a large excess of random passenger events. We propose a non-parametric, robust analytical framework to identify such driving disruptions that unifies three popular methods used for expression outlier detection. We extend these methods to incorporate an empirically-based cutoff based upon a change point model, as opposed to requiring an a priori specified cutoff that does not take into account distribution differences among genes. We compare results based on this extension, and illustrate our approach using available public domain data on various cancer types. We further discuss the clinical significance of defined outlier genes and the application of our approach as a filter applied to translocations obtained from NGS data.