The effect of sample splitting and technical replication on mRNA-seq and sRNA-seq quantification

RNA sequencing (RNA-seq) is a widely used method for quantifying RNA levels across the environmental, biological and medical sciences. The accuracy of the output from an RNA-seq experiment is known to vary due to sequencing biases or errors in quantification. These have the potential to lead to false calls of differential expression (DE), and hence, to affect the accuracy of the biological inference. A proposed solution to reduce the number of false positives and increase confidence in the quality of results from such experiments is to increase the number of biological replicates. In addition, more recent suggestions are to create additional technical replicates within biological replicates (i.e. to split samples across sequencing lanes). The optimal strategy for analysing and normalizing such data, and for maximising accuracy as a function of cost, biological and technical replication is important to understand, yet currently unclear. The aim of this study was to test the effect of technical replication and sample splitting on the overall outcome of gene expression profiling for RNA-seq data. Overall design: To compare the effect of introducing technical replicates, and as a result, reducing the sequencing depth allocated per sample, we sequenced 3 samples both full and split approaches. The full approach is equivalent to sequencing the whole sample on one lane. The split approach is equivalent to creating two technical replicates, 50% each. The full samples corresponding to this study are presented in GSE55839. The samples are from wild type Dahomey D. melanogaster males that were exposed to rival males ('R') for 2 or 26 hours and split into Head-Thorax 'H' or Abdomen 'A' samples. Males were reared under standardised conditions and snap frozen at the same time of day. The mRNA-seq was single end, 50nt reads, standard Illumina adapters and using the HiSeq-2500 platform. Samples 2RA3_1 and 2 (2h, Rivals, Abdomen) are samples of the same biological sample that were split and sequenced on different lanes of the same flow cell, 2RH2_1 and 2 similarly for the 2h, Rivals, Head-Thorax sample and 26RH3_1 and 2 similarly for the 26h Rivals Head-Thorax sample.