Cancer Diagnosis Program | Biorepositories and Biospecimen Research Branch

Search BRD Papers

Small RNA-sequencing for Analysis of Circulating miRNAs: Benchmark Study.

Author(s): Androvic P, Benesova S, Rohlova E, Kubista M, Valihrach L

Publication: J Mol Diagn, 2022, Vol. , Page

PubMed ID: 35081459 PubMed Review Paper? No

Purpose of Paper

Conclusion of Paper

Studies

1. Study Purpose

   This study assessed the potential bias introduced by small RNA sequencing library preparation methods by comparing the next generation sequencing results of the miRXplore universal reference with pooled plasma using 8 different library preparation methods. K₂EDTA blood was collected from three healthy volunteers and plasma was separated by centrifugation at 1500 g for 15 min followed by 3000 g for 15 min, within 30 min of collection. Plasma was stored at -80°C until RNA extraction. RNA was extracted from pooled plasma aliquots using the miRNeasy Serum/Plasma Advanced Kit and evaluated using a real-time PCR panel. Sequencing libraries were prepared in duplicate from RNA extracted from plasma pools and the miRXplore Universal Reference using Norgen, Lexogen, QIAseq (with and without deduplication), NEXTflex, RealSeq, SMARTer, and EdgeSeq. Library yield was quantified by Qubit and fragment size was analyzed by an Agilent Fragment analyzer. Pooled libraries generated with each protocol were separated on a 5% TBE-PAGE on a Mini-PROTEAN tetra cell and fragments of 140-160 nt were cut out and purified with SPRIselect reagent. All libraries were sequenced on a NextSeq 500 high-output machine. EdgeSeq libraries were sequenced on a TATAA Biocenter machine. To verify sequencing results, 35 selected miRNAs were quantified in plasma by real-time PCR..

   Summary of Findings:

   The greatest percentage of miRNA that fell within 2-fold of expected levels was achieved when the universal reference was sequenced using libraries constructed with EdgeSeq (83%), followed by SMARTer (51%), NextFlex (32%), QIAseq with deduplication (UMI) (30%), QIAseq (29%), RealSeq (21%), Lexogen (14%) and Norgen (13%). Further evaluation using the reference miRNA and RealSeq revealed that most bias was attributed to the ligation method and the least to PCR. Further, the first nucleotide contributed to 44% of the variability with RealSeq and 25% of the variability with SMARTer; while with Lexagen, Norgen and QIAseq 8-10% of variability was attributed to last two bases (and 0.2-19.2% of variability to the free energy of the adapter-miRNA complex (, further supporting ligation as the main source of bias.  There were also protocol-specific 3’, 5’, and base bias in the detection of false isomiRs, with SMARTer libraries resulting in the highest percentage of false isomiR detection (4% versus <0.4% for all other methods) using the reference miRNA.

   Mapping rate was highly dependent on library preparation method, as libraries prepared with SMARTer had a very low mapping rate (<10%) while those prepared with EdgeSeq had a very high mapping rate (95%). For all library preparation methods, the majority of mapped reads mapped to the most abundant miRNA, with a single miRNA accounting for >50% of all mapped reads for most methods. Importantly, the top 10 ranked miRNA differed among library preparation methods, although miR-451 and miR-16 appeared on all lists. When miRNAs were simply classified as present or absent, 5 million reads was sufficient to reach saturation for all library preparation methods with the exception of RealSeq. The highest number of miRNAs were detected using EdgeSeq libraries which allowed detection of hundreds of more miRNAs than were detected using any of the other libraries. The authors attribute the elevated number of miRNAs detected in EdgeSeq libraries to a higher mapping rate and a higher false positive rate (lower specificity) compared to other libraries. The correlation between actual levels of miRNA in plasma (determined by real-time PCR) and those determined by sequencing ranged from R²=0.53 for Norgen to R²=0.83 for SMARTer; however, when corrected for protocol-specific bias (determined using the reference miRNA) the strength of the correlation between actual miRNA levels and those determined by NGS increased significantly (P=1.06 x 10^-6)  for each  library preparation method with the exception of EdgeSeq (R²=0.8 without correction and 0.78 after correction). Correction for bias also increased the strength of the correlation in miRNA-Seq data between library preparation methods (without correction R²=0.66-1.0 and after correction R²=0.79-1.0).

   Biospecimens

   • Bodily Fluid - Plasma

   Preservative Types

   • Frozen

   Diagnoses:

   • Normal

   Platform:

   Analyte	Technology Platform
   RNA	Next generation sequencing
   RNA	Spectrophotometry
   RNA	Real-time qRT-PCR

   Pre-analytical Factors:

   Classification	Pre-analytical Factor	Value(s)
   Next generation sequencing Specific	Priming method	Norgen library Lexogen library QIAseq library QIAseq library with deduplication (UMI) NEXTflex library RealSeq library SMARTer library EdgeSeq library

   View more