miRNA Temporal Analyzer (mirnaTA): a bioinformatics tool for identifying differentially expressed microRNAs in temporal studies using normal quantile transformation

• Published in: Gigascience 2014-10, Vol.3 (1), p.20-20
• Main Authors: Cer, Regina Z, Herrera-Galeano, J Enrique, Anderson, Joseph J, Bishop-Lilly, Kimberly A, Mokashi, Vishwesh P
• Format: Article
• Language: English
• Subjects: Bioinformatics; Cluster analysis; Datasets; Distribution functions; Euclidean geometry; Medical research; MicroRNAs; miRNA; Next-generation sequencing; Open source software; Pattern analysis; Quantiles; Regression models; Scientific visualization; Scientists; Statistical analysis; Technical Note; Transformations (mathematics); User requirements; Variance analysis
• ISSN: 2047-217X; 2047-217X
• DOI: 10.1186/2047-217X-3-20

Abstract Background Understanding the biological roles of microRNAs (miRNAs) is a an active area of research that has produced a surge of publications in PubMed, particularly in cancer research. Along with this increasing interest, many open-source bioinformatics tools to identify existing and/or discover novel miRNAs in next-generation sequencing (NGS) reads become available. While miRNA identification and discovery tools are significantly improved, the development of miRNA differential expression analysis tools, especially in temporal studies, remains substantially challenging. Further, the installation of currently available software is non-trivial and steps of testing with example datasets, trying with one's own dataset, and interpreting the results require notable expertise and time. Subsequently, there is a strong need for a tool that allows scientists to normalize raw data, perform statistical analyses, and provide intuitive results without having to invest significant efforts. Findings We have developed miRNA Temporal Analyzer (mirnaTA), a bioinformatics package to identify differentially expressed miRNAs in temporal studies. mirnaTA is written in Perl and R (Version 2.13.0 or later) and can be run across multiple platforms, such as Linux, Mac and Windows. In the current version, mirnaTA requires users to provide a simple, tab-delimited, matrix file containing miRNA name and count data from a minimum of two to a maximum of 20 time points and three replicates. To recalibrate data and remove technical variability, raw data is normalized using Normal Quantile Transformation (NQT), and linear regression model is used to locate any miRNAs which are differentially expressed in a linear pattern. Subsequently, remaining miRNAs which do not fit a linear model are further analyzed in two different non-linear methods 1) cumulative distribution function (CDF) or 2) analysis of variances (ANOVA). After both linear and non-linear analyses are completed, statistically significant miRNAs (P?