Analysis of RNA-Seq Data from Complex Factorial Experiments

RNA-Seq is now used routinely to investigate differential gene expression, and many methods are available to analyse simple experiments which compare two samples (treated and control).  However, biological processes often depend on several factors, and require experiments with more complex designs.  These experiments need methods to analyse RNA-Seq data which make full use of factorial design to identify differential gene expression associated with individual factors and their interactions.

We present a method to analyse RNA-Seq data from a complex factorial experiment, designed to investigate gene expression in an elite biofuel willow subjected to reaction wood induction (tipping at 45°).  The experimental factors included tipping treatments and tissue type, with samples taken at 5 time points after tipping.  The RNA-Seq data were analysed within the generalized linear model framework, using analysis of deviance to identify differentially expressed genes in distinct groups.  The gene groups were related to precise biological questions based on the structure of the treatment model.

RNA-Seq data from complex, designed factorial experiments can be analysed using established methods within the framework of generalized linear models.  This is more experimentally efficient and statistically powerful than conducting pairwise tests of single factors.  Alternative models can be formulated to answer different, specific biological questions of interest.  The same modelling framework could be used to analyse yields and other agronomic data from the same experiment, allowing more comprehensive interpretation of the combined data.  These methods extend the range of experiments amenable to RNA-Seq data analysis.