Abstract
Currently, quantitative RNA-seq methods are pushed to work with increasingly small starting amounts of RNA that require amplification. However, it is unclear how much noise or bias amplification introduces and how this affects precision and accuracy of RNA quantification. To assess the effects of amplification, reads that originated from the same RNA molecule (PCR-duplicates) need to be identified. Computationally, read duplicates are defined by their mapping position, which does not distinguish PCR- from natural duplicates and hence it is unclear how to treat duplicated reads. Here, we generate and analyse RNA-seq data sets prepared using three different protocols (Smart-Seq, TruSeq and UMI-seq). We find that a large fraction of computationally identified read duplicates are not PCR duplicates and can be explained by sampling and fragmentation bias. Consequently, the computational removal of duplicates does improve neither accuracy nor precision and can actually worsen the power and the False Discovery Rate (FDR) for differential gene expression. Even when duplicates are experimentally identified by unique molecular identifiers (UMIs), power and FDR are only mildly improved. However, the pooling of samples as made possible by the early barcoding of the UMI-protocol leads to an appreciable increase in the power to detect differentially expressed genes.
Highlights
High throughput RNA sequencing methods (RNA-seq) are currently replacing microarrays as the method of choice for gene expression quantification[1,2,3,4,5]
We investigate the source of read duplicates by analysing PCR bias and fragmentation bias, assess the accuracy using ERCCs - spike-in mRNAs of known concentrations17,18 - and assess precision using power simulations using PROPER19
We started with two benchmarking datasets from the literature[2] that sequenced five replicates of bulk mRNA using the TruSeq protocol on commercially available reference mRNAs: the Universal Human Reference RNA (UHRR; Agilent Technologies) and the Human Brain Reference RNA (HBRR, ThermoFisher Scientific)
Summary
High throughput RNA sequencing methods (RNA-seq) are currently replacing microarrays as the method of choice for gene expression quantification[1,2,3,4,5]. Methods using UMIs and cellular barcodes have been optimized for low starting amounts and low costs, to generate RNA-seq profiles from single cells[7,15] To achieve these goals, the methods differ in a number of steps that will impact the probability of read duplicates and their detection (Fig. 1). The chance that such natural duplicates, i.e. read duplicates that originated from different mRNA molecules, occur for a transcript of a given length, increases with expression levels and fragmentation bias That said, it is unclear whether removing read duplicates computationally improves accuracy and precision by reducing PCR bias and noise or whether it decreases accuracy and precision by removing genuine information. We investigate the source of read duplicates by analysing PCR bias and fragmentation bias, assess the accuracy using ERCCs - spike-in mRNAs of known concentrations17,18 - and assess precision using power simulations using PROPER19
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
