RNaseIII and T4 Polynucleotide Kinase sequence biases and solutions during RNA-seq library construction

Abstract

BackgroundRNA-seq is a next generation sequencing method with a wide range of applications including single nucleotide polymorphism (SNP) detection, splice junction identification, and gene expression level measurement. However, the RNA-seq sequence data can be biased during library constructions resulting in incorrect data for SNP, splice junction, and gene expression studies. Here, we developed new library preparation methods to limit such biases.ResultsA whole transcriptome library prepared for the SOLiD system displayed numerous read duplications (pile-ups) and gaps in known exons. The pile-ups and gaps of the whole transcriptome library caused a loss of SNP and splice junction information and reduced the quality of gene expression results. Further, we found clear sequence biases for both 5' and 3' end reads in the whole transcriptome library. To remove this bias, RNaseIII fragmentation was replaced with heat fragmentation. For adaptor ligation, T4 Polynucleotide Kinase (T4PNK) was used following heat fragmentation. However, its kinase and phosphatase activities introduced additional sequence biases. To minimize them, we used OptiKinase before T4PNK. Our study further revealed the specific target sequences of RNaseIII and T4PNK.ConclusionsOur results suggest that the heat fragmentation removed the RNaseIII sequence bias and significantly reduced the pile-ups and gaps. OptiKinase minimized the T4PNK sequence biases and removed most of the remaining pile-ups and gaps, thus maximizing the quality of RNA-seq data.ReviewersThis article was reviewed by Dr. A. Kolodziejczyk (nominated by Dr. Sarah Teichmann), Dr. Eugene Koonin, and Dr. Christoph Adami. For the full reviews, see the Reviewers' Comments section.

Phospho‐RNA‐seq: a modified small RNA‐seq method that reveals circulating mRNA and lncRNA fragments as potential biomarkers in human plasma

“…Although PNK treatment enabled much greater recovery of sequences lacking 5′P and/or 3′OH, PNK is known to exhibit target nucleic acid sequence‐dependent biases in its kinase and phosphatase activities (Lee et al , ). These very likely influence the distribution of sequences we are able to recover with the current iteration of the phospho‐RNA‐seq method, limiting the accuracy of estimating the relative abundance of different mRNA/lncRNA fragments within a sample.…”

“…Our incorporation of synthetic RNA pools as a ground truth reference was especially important in our study, as it allowed us to demonstrate clearly that lack of a 5 0 P and presence of a 3 0 P are both impediments to recovery of these fragments by standard small RNA-seq. Although PNK treatment enabled much greater recovery of sequences lacking 5 0 P and/or 3 0 OH, PNK is known to exhibit target nucleic acid sequence-dependent biases in its kinase and phosphatase activities (Lee et al, 2013). These very likely influence the distribution of sequences we are able to recover with the current iteration of the phospho-RNA-seq method, limiting the accuracy of estimating the relative abundance of different mRNA/lncRNA fragments within a sample.…”

RNA-seq of human reference RNA samples using a thermostable group II intron reverse transcriptase

“…Such biases are well documented for random hexamer priming used for cDNA synthesis in both TruSeq methods (Hansen et al 2010), but could also arise from A-addition to the 3 ′ ends of dsDNAs prior to adaptor ligation or the ligation itself in TruSeq, or from the template-switching reaction used for RNA-seq adapter attachment in TGIRT-seq. In addition, 5 ′ and 3 ′ biases for all methods could arise from other steps, including nontemplated nucleotide addition at the 3 ′ ends of cDNAs after reaching the 5 ′ end of the RNA template, RNA fragmentation, dephosphorylation, or DNA or RNA ligases used to attach RNA-seq adapters (Hafner et al 2011;Kwok et al 2013;Lee et al 2013;Wery et al 2013;Zajac et al 2013;Jackson et al 2014). As the TruSeq libraries were generated by using random hexamer priming for cDNA synthesis, biases at both the 5 ′ and 3 ′ ends of the RNA reads reflect where the primer anneals and may include sequence errors due to mispriming.…”

Digital Gene Expression Analysis with Sample Multiplexing and PCR Duplicate Detection: A Straightforward Protocol

“…Methods known to have less bias in RNA library preparation were used, while different steps were coupled as much as possible to avoid unnecessary clean-up steps. RNA fragmentation was carried out by heating, which was shown previously to be superior to RNase treatment (32). Direct P5 adapter RNA ligation was performed right after the fragmentation step, which is one of the least biased approaches among RNA library preparation methods (13) and also eliminates the need for a second-strand synthesis step before PCR.…”