Statistical Methods for the Analysis of Next-Generation Sequencing Data

• Activity Number: 72
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Karthik Devarajan*+
• Companies: Fox Chase Cancer Center
• Keywords: next-generation sequencing ; digital gene expression ; nonnegative matrix factorization ; high-dimensional data ; ChIP-Seq, RNA-Seq, SAGE ; over-dispersion

Abstract

The emergence of next-generation sequencing technologies such as RNA-Seq have resulted in massive amounts of digital gene expression data in the form of millions of genome-wide sequence reads. These data represent the sequence tag counts across the genome for each biological sample and typically exhibit over-dispersion across samples. One of the objectives of the investigator is to identify differentially expressed tags between two or more experimental conditions.

The purpose of this paper is two-fold. First, we survey the current literature in this area and compare the performance of existing methods for the analysis of digital gene expression data. Next, we investigate the application of nonnegative matrix decompositions for handling count data as an exploratory tool for dimension reduction, visualization and interpretation. In this high-dimensional setting, we propose an approach for handling over-dispersion in the tag counts and illustrate the utility of our methods using real-life and simulated digital gene expression data. Our methodology is also applicable to data arising from other similar technologies such as ChIP-Seq and SAGE.