Differentially expressed genes from RNA-Seq and functional enrichment results are affected by the choice of single-end versus paired-end reads and stranded versus non-stranded protocols

We undertook four mammalian transcriptomics experiments to compare the effect of read mapping, feature counting and differential expression analysis using single-end (SE) and paired-end (PE) protocols. For three of these experiments we also compared a non-stranded (NS) and a strand-specific approach to mapping the paired-end data. Overall design: Each of the four RNA-Seq experiments had a simple but typical design, comprised of three biological controls and three treated samples. The six samples in each of the experiments were independent biological replicates. Experiments 1 and 2 were from mouse tissue or primary cells whereas Experiments 3 and 4 involved human primary cells and a cell line respectively. In all four experiments sequencing was performed on both ends of the cDNA fragment (paired-end reads). We mapped both ends (see Methods) to produce our paired-end data set (PE data). We also mapped only the first read to produce our single-end read datasets for each experiment (SE data). Sequencing for Experiment 1 occurred in 2012 and used a non-strand-specific protocol for library preparation (Illumina TruSeq kit). The other three experiments were sequenced more recently with the Illumina TruSeq Standed library preparation kit [13]. For these three experiments we looked at the effect of analyzing the paired-end data with a protocol that recognizes the strand-specific nature of the reads (PE data) and also with a protocol that does not recognize this (NS data). In this way we could assess the difference that a strand-specific protocol makes to gene expression analysis.