Abstract
RNA sequencing (RNA-seq) is currently the standard method for genome-wide expression profiling. RNA-seq reads often need to be mapped to a reference genome before read counts can be produced for genes. Read trimming methods have been developed to assist read mapping by removing adapter sequences and low-sequencing-quality bases. It is however unclear what is the impact of read trimming on the quantification of RNA-seq data, an important task in RNA-seq data analysis. In this study, we used a benchmark RNA-seq dataset and simulation data to assess the impact of read trimming on mapping and quantification of RNA-seq reads. We found that adapter sequences can be effectively removed by read aligner via 'soft-clipping' and that many low-sequencing-quality bases, which would be removed by read trimming tools, were rescued by the aligner. Accuracy of gene expression quantification from using untrimmed reads was found to be comparable to or slightly better than that from using trimmed reads, based on Pearson correlation with reverse transcriptase-polymerase chain reaction data and simulation truth. Total data analysis time was reduced by up to an order of magnitude when read trimming was not performed. Our study suggests that read trimming is a redundant process in the quantification of RNA-seq expression data.