Abstract
Although a large number of clustering algorithms have been proposed to identify groups of co-expressed genes from microarray data, the question of if and how such methods may be applied to RNA-seq data remains unaddressed. In this work, we investigate the use of data transformations in conjunction with Gaussian mixture models for RNA-seq co-expression analyses, as well as a penalized model selection criterion to select both an appropriate transformation and number of clusters present in the data. This approach has the advantage of accounting for per-cluster correlation structures among samples, which can be quite strong in RNA-seq data. In addition, it provides a rigorous statistical framework for parameter estimation, an objective assessment of data transformations and number of clusters, and the possibility of performing diagnostic checks on the quality and homogeneity of the identified clusters. We analyze four varied RNA-seq datasets to illustrate the use of transformations and model selection in conjunction with Gaussian mixture models. Finally, we propose an R package coseq (co-expression of RNA-seq data) to facilitate implementation and visualization of the recommended RNA-seq co-expression analyses.
Footnotes
Andrea Rau is a Research Scientist in the Animal Genetics and Integrative Biology research unit at the French National Institute for Agronomical Research (INRA) in Jouy en Josas, France. Her research focuses on developing statistical methodology and open-source software for the analysis of genomic and transcriptomic data.
Cathy Maugis-Rabusseau is an Associate Professor at the French National Institute of Applied Sciences and the French Institute of Mathematics in Toulouse, France. Her research is centered on theoretical and methodological developments for model-based clustering methods, variable selection, and hypothesis testing approaches.