Statistical methods for identifying differentially expressed genes in RNA-Seq experiments

Transcriptomics holds the key to understanding how the information encoded in the genome is translated into cellular functions, and how this translation process responds to the changing environment. Given a transcriptome, or the collection of all the transcripts including both protein coding mRNAs and noncoding RNAs, one of the outstanding questions in transcriptomics is to accurately quantify the abundance of each transcript within different tissues and time points, and to correlate changes in abundance to genetic and environmental perturbation in order to understand genome function and adaptation.Transcriptome profiling, or gene expression profiling, is the technology used to determine the steady state abundance of each transcript within a transcriptome. Transcriptome profiling is traditionally done using either quantitative RT-PCR (reviewed in [1]) to interrogate a few genes, or microarrays (cDNA array [2] or whole genome tiling array [3-5]) to investigate genome-wide transcriptional activity. Recently, as a result of the low cost of next generation sequencing technologies [6], transcriptome profiling by RNA sequencing (RNA-Seq) is becoming the method of choice because of its unprecedented sensitivity and accuracy (reviewed in [7,8]). Unlike prior technologies, next-generation sequencing technologies allow reference transcriptomes to be assembled directly from RNA-Seq data, thereby eliminating the need for existing reference genomes or transcriptomes [9]. This capability is particularly attractive for non-model organisms or microbial communities that lack high quality references.There are both shared and unique aspects in the experimental design and data generation phases of expression microarrays and RNA-Seq technologies, and these attributes are compared elsewhere [7,8,10,11]. For data analysis there are three major steps for both technologies: data preprocessing, statistical analysis and functional interpretation (Figure？1). Preprocessing microarray data norma