Fig. 1

Study Design. a We evaluated the influence of starting amount on the consistency of results, as well as the accuracy of results obtained when using a variety of methods, including those intended to reduce bias from adapter ligation, reverse transcription (RT), and PCR amplification. b We compared four commercially available kits and two preprocessing methods to address RT and PCR bias. c In the Deduped method we collapsed duplicate reads based on a unique molecular identifier (UMI) of degenerate bases in the adapter sequences (bases within the black boxes). We also compared the collapsed data with a 5% subset of the NEXTflex data to determine if performance differences were due to the UMI-based collapsing of reads or simply due to having fewer reads. d We evaluated two data types: miRNA quantifications from homogenate whole brain total RNA and miRNA quantifications from a pool of 962 equimolar synthetic RNAs with sequences corresponding to human, rat, mouse, and virus miRNA. We had two batches of human brain data. The first included triplicates of different starting amounts based on the kit manufacturers’ suggested ranges. The second included a single sample of the same human brain with 1000 ng of input. We used 300 ng of the synthetic miRNAs for each tested method. e Our processing pipelines for the two types of RNA studied. f We evaluated the 6 small-RNA sequencing methods using 4 major assessments. The brain icon indicates utilization of brain samples to assess a question, while the red tube indicates utilization of synthetic miRNA samples