Adapter dimer contamination in sRNA‐sequencing datasets predicts sequencing failure and batch effects and hampers extracellular vesicle‐sRNA analysis

Abstract

AbstractSmall RNA (sRNA) profiling of Extracellular Vesicles (EVs) by Next‐Generation Sequencing (NGS) often delivers poor outcomes, independently of reagents, platforms or pipelines used, which contributes to poor reproducibility of studies. Here we analysed pre/post‐sequencing quality controls (QC) to predict issues potentially biasing biological sRNA‐sequencing results from purified human milk EVs, human and mouse EV‐enriched plasma and human paraffin‐embedded tissues. Although different RNA isolation protocols and NGS platforms were used in these experiments, all datasets had samples characterized by a marked removal of reads after pre‐processing. The extent of read loss between individual samples within a dataset did not correlate with isolated RNA quantity or sequenced base quality. Rather, cDNA electropherograms revealed the presence of a constant peak whose intensity correlated with the degree of read loss and, remarkably, with the percentage of adapter dimers, which were found to be overrepresented sequences in high read‐loss samples. The analysis through a QC pipeline, which allowed us to monitor quality parameters in a step‐by‐step manner, provided compelling evidence that adapter dimer contamination was the main factor causing batch effects. We concluded this study by summarising peer‐reviewed published workflows that perform consistently well in avoiding adapter dimer contamination towards a greater likelihood of sequencing success.