PCR clonal artefacts originating from NGS library preparation can affect both genomic as well as RNA-Seq applications when protocols are pushed to their limits. In RNA-Seq however the artifactual reads are not easy to tell apart from normal read duplication due to natural over-sequencing of highly expressed genes. Especially when working with little input material or single cells assessing the fraction of duplicate reads is an important quality control step for NGS data sets. Up to now there are only tools to calculate the global duplication rates that do not take into account the effect of gene expression levels which leaves them of limited use for RNA-Seq data.
Researchers at the Institute of Molecular Biology, Mainz, Germany have devloped the tool dupRadar, which provides an easy means to distinguish artefactual from natural duplicate reads in RNA-Seq data. dupRadar assesses the fraction of duplicate reads per gene dependent on the expression level. Apart from the Bioconductor package dupRadar the developers provide shell scripts for easy integration into processing pipelines.
Several RNA-seq datasets from Marinov et al. Legends shows the intercept and slope of a fitted logit model. A Single cell experiment with relatively low duplication rates and most of the genes detected. B Single cell experiment with most of the genes undetected and high duplication rate on the detected ones. C RNA-seq experiment pushing the protocol to only 100pg of input material, with low duplication rates and relatively good identification of genes. D same RNA-seq experiment, showing over-sequencing due to higher sequencing depth of the library.
The Bioconductor package dupRadar offers straight-forward methods to assess RNA-Seq datasets for quality issues with PCR duplicates. It is aimed towards simple integration into standard analysis pipelines as a default QC metric that is especially useful for low-input and single cell RNA-Seq data sets.
Availability – http://bioconductor.org/packages/dupRadar/