Abstract
Advances in high-throughput sequencing have enabled profiling of microRNAs (miRNAs), however, a consensus pipeline for sequencing of small RNAs has not been established. We built and optimized an analysis pipeline using Partek Flow, circumventing the need for analyzing data via scripting languages. Our analysis assessed the effect of alignment reference, normalization method, and statistical model choice on biological data. The pipeline was evaluated using sequencing data from HaCaT cells transfected with either a non-silencing control or siRNA against ΔNp63α, a p53 family member protein which is highly expressed in non-melanoma skin cancer and shown to regulate a number of miRNAs. We posit that 1) alignment and quantification to the miRBase reference provides the most robust quantitation of miRNAs, 2) normalizing sample reads via Trimmed Mean of M-values is the most robust method for accurate downstream analyses, and 3) use of the lognormal with shrinkage statistical model effectively identifies differentially expressed miRNAs. Using our pipeline, we identified previously unrecognized regulation of miRs-149-5p, 18a-5p, 19b-1-5p, 20a-5p, 590-5p, 744-5p and 93-5p by ΔNp63α. Regulation of these miRNAs was validated by RT-qPCR, substantiating our small RNA-Seq pipeline. Further analysis of these miRNAs may provide insight into ΔNp63α's role in cancer progression. By defining the optimal alignment reference, normalization method, and statistical model for analysis of miRNA sequencing data, we have established an analysis pipeline that may be carried out in Partek Flow or at the command line. In this manner, our pipeline circumvents some of the major hurdles encountered during small RNA-Seq analysis.