Comprehensive assessment of multiple biases in small RNA sequencing reveals significant differences in the performance of widely used methods

Carrie Wright,Martha Kimos,Joo Heon Shin,Emily E Burke,Courtney Williams,Anandita Rajpurohit,Alan J Cross,Daniel R Weinberger,Leonardo Collado-Torres,Andrew E Jaffe,Nicholas J Brandon

doi:10.1186/s12864-019-5870-3

Abstract

BackgroundRNA sequencing offers advantages over other quantification methods for microRNA (miRNA), yet numerous biases make reliable quantification challenging. Previous evaluations of these biases have focused on adapter ligation bias with limited evaluation of reverse transcription bias or amplification bias. Furthermore, evaluations of the quantification of isomiRs (miRNA isoforms) or the influence of starting amount on performance have been very limited. No study had yet evaluated the quantification of isomiRs of altered length or compared the consistency of results derived from multiple moderate starting inputs. We therefore evaluated quantifications of miRNA and isomiRs using four library preparation kits, with various starting amounts, as well as quantifications following removal of duplicate reads using unique molecular identifiers (UMIs) to mitigate reverse transcription and amplification biases.ResultsAll methods resulted in false isomiR detection; however, the adapter-free method tested was especially prone to false isomiR detection. We demonstrate that using UMIs improves accuracy and we provide a guide for input amounts to improve consistency.ConclusionsOur data show differences and limitations of current methods, thus raising concerns about the validity of quantification of miRNA and isomiRs across studies. We advocate for the use of UMIs to improve accuracy and reliability of miRNA quantifications.

Highlights

RNA sequencing offers advantages over other quantification methods for microRNA, yet numerous biases make reliable quantification challenging
Study design In this study we evaluated the influence of several potential sources of bias and inconsistency on miRNA quantifications (Fig. 1a) by comparing the performance of four commercially available kits including two that are designed to mitigate adapter ligation bias in different ways (Fig. 1b) and two preprocessing methods including one to mitigate reverse transcription and adapter ligation bias and a control for comparison (Fig. 1c), as well as various starting amounts (100 ng to 2000 ng) for each method to determine the reliability of results achieved with smaller inputs (Fig. 1d)
We evaluated the influence of reverse transcription and amplification bias by utilizing the random sequences within the adapters of the NEXTflex kit, as unique molecular identifiers (UMIs)

Summary

Introduction

RNA sequencing offers advantages over other quantification methods for microRNA (miRNA), yet numerous biases make reliable quantification challenging. Small RNA sequencing (sRNA-seq) allows for detection of novel miRNAs and altered canonical miRNA sequences, termed isomiRs [5,6,7] These miRNA isoforms are produced by many mechanisms, Wright et al BMC Genomics (2019) 20:513 assessment of the quantification of isomiRs using sRNA-seq; and miRNA quantifications using this method are often inconsistent across studies [16]. This is likely in part due to differences between methods and/or variation in the detection by individual methods [17] (from library preparation to preprocessing to normalization, etc.). Evaluations and comparisons of the accuracy (how close measurements are to the truth) and consistency (how close measurements are across replicates) associated with current methods are critical for proper cross-study interpretation and for guiding methodological improvement

Methods

Results

Discussion

Conclusion