DGEclust: differential expression analysis of clustered count data

TL;DR

DGEclust is a novel statistical method for clustering count data from sequencing experiments that simultaneously determines the number of clusters and estimates parameters, improving differential expression analysis accuracy.

Contribution

It introduces a unified approach for clustering and differential expression analysis of count data, addressing model selection and uncertainty in parameter estimation.

Findings

Outperforms popular alternative methods in accuracy

Supports a broader class of problems

Provides a unified framework for clustering and differential analysis

Abstract

Most published studies on the statistical analysis of count data generated by next-generation sequencing technologies have paid surprisingly little attention on cluster analysis. We present a statistical methodology (DGEclust) for clustering digital expression data, which (contrary to alternative methods) simultaneously addresses the problem of model selection (i.e. how many clusters are supported by the data) and uncertainty in parameter estimation. We show how this methodology can be utilised in differential expression analysis and we demonstrate its applicability on a more general class of problems and higher accuracy, when compared to popular alternatives. DGEclust is freely available at https://bitbucket.org/DimitrisVavoulis/dgeclust