MiRNA-Seq normalization comparisons need improvement

Abstract

As developers and users of informatics strategies, we are keenly interested in the relative merits of competing approaches. Crucially, there has been relatively little investigation into normalization strategies for miRNA-Seq data and the timely article from Garmire and Subramaniam promised to shed light on this issue. Unfortunately, errors in the implementation, poor choice of performance metrics (or poor choice of data set), few details about practical implementation (e.g., elimination of features containing zero count), and sensitivity to choices made regarding the reference truth data set have left many open questions about the best analysis methods for miRNA-Seq data. In this paper, we ..