Assessing agreement between microRNA microarray platforms via linear measurement error models

Over the last years miRNA microarray platforms have provided insights in the biological mechanisms underlying onset and development of several diseases and have thus become a very popular instrument for profiling thousands of miRNA simultaneously. However, because of large variety of microarray platforms available, an assessment of their performance in terms of both within-platform reliability and between-platform agreement is needful. In particular, assessment of platform concordance has been a very relevant issue in the past decade. To date, only a few studies have evaluated this problem in the field of miRNA microarray, and mostly by using improper statistical methods such as the Pearson and Spearman correlation coefficients. In this work we suggest to use a recently proposed modified version of the classical Bland-Altman approach for comparing clinical measurement methods. This modified version is useful in that allows not only to evaluate agreement between different miRNA microarray platforms, but also to assess which are the potential sources of disagreement/bias between them. Two samples were profiled using Affymetrix, Agilent and Illumina miRNA platform using three technical replicates each, and pairwise agreement between platforms was evaluated within each sample. Our results suggest that, after bias correction, Illumina and Agilent show the best patterns of agreement for both samples involved in the experiment, whereas Affymetrix is the one which seem to "disagree" most, suggesting that a linear relationship as that hypothesized by the measurement error model used is not able to capture the complexity of the phenomenon. In the future it will be interesting to apply this method also to the comparison of microarray and NGS platform, a topic which is becoming more and more relevant, also by adopting non-linear measurement error models to depict relationships between platforms.