1. Introduction
MiRNAs are small non-coding RNA molecules that have been shown to play a critical role in tumorigenesis [1,2,3,4] and in several other pathologies [5,6,7,8]. In order to measure miRNA intensity levels, several methods, such as RT-qPCR, high-throughput sequencing and microarrays, have been developed and have enabled researchers to profile a large number of miRNAs simultaneously across different experimental conditions [9]. MiRNA microarrays, in particular, since their first appearance in 2004 [10], have known a considerable expansion in life sciences and are now routinely used in biomolecular research.
In the last few years, miRNA microarrays have been compared with next generation sequencing technologies to study their performances [11,12]. The main interest lies in the potentialities and advantages offered by these new platforms in terms, for instance, of new miRNAs discovery [13]. However, since results from these comparison studies appear to be contrasting, miRNA microarrays still remain an effective and useful technology, whose characteristics need to be properly assessed, both in terms of within-platform reliability and of between-platforms agreement.
To date, only a few studies have attempted to evaluate within-platform [14] and between-platform [15,16] reliability in miRNA microarrays. Results were reported mainly as correlation coefficients (both Pearson and Spearman) for evaluating both intra-platform and inter-platform performance, calculated on a subset of miRNA that depended on detection calls concordant between platforms (i.e., miRNA that were called “detected/present” on all platforms considered for the analysis). Additionally, Sato and colleagues assessed between-platform comparability also in terms of miRNAs that were commonly differentially expressed between samples for all platforms [15], whereas Yauk et al. [16] evaluated within-platform reproducibility via Lin’s concordance correlation coefficient [17].
In this work, a different approach based on the Bland–Altman method to assess between-platform agreement is proposed. In particular, the proposed method is applied to assess agreement between three different miRNA microarray platforms (Affymetrix, Agilent, Illumina). Additionally, advice against the use of Pearson/Spearman correlation coefficients is provided, and use of concordance correlation coefficients (pairwise and overall) is suggested as a better measure to evaluate within-platform reliability.