Comparison of RNA-seq Quantification Software

RNA-seq is a more accurate technology in measuring transcripts levels by using high-throughput sequencing of cDNA. However, the quantification of mRNA abundances from RNA-seq data may be biased due to various biological or statistical effects. Several RNA-seq quantification software, including RSEM, Cufflinks, IsoEM, Genominator, and RNASeqBias, had been recently proposed to correct such biases. The objective of this study was to compare the above five software by applying RNA-seq analysis to a benchmark MAQC human brain data while the Taqman qRT-PCR dataset was treated as the golden-standard to evaluate them. Different software was compared to each other based on their associations between the log expression values that obtained from each method and the Taqman log value. In addition, we also discussed the level of biases correction for the programs. According to the mapping results, it was observed that the transcript length effects did exist in the MAQC data. The analytic results showed that all software can reduce length biases. Although Cufflinks, IsoEM, Genominator, and RNASeqBias have the functions to correct sequence-specific biases, only Cufflinks has a bit apparent to correct sequence-specific biases. In conclusion, Cufflinks has the best performance in biases correction for RNA-seq data.