Reliability of miRNA microarray platforms: An approach based on random effects linear models

MiRNAs are short ribonucleic acid (RNA) molecules, acting as post-transcriptional regulators. Intensity levels of thousand of miRNAs are commonly measured via microarray platforms,with pros and cons similar to those for gene expression arrays. Data reliability for miRNA microarrays is a crucial point to obtain correct estimates of miRNA intensity, and maximizing biological relative to technical variability is a task that has to be properly addressed. To such aim, random effects models provide a powerful instrument to characterize different sources of variability. Here we evaluated repeatability of Affymetrix Gene Chip © miRNA Array by fitting random effects models separately for 4 cell lines. Results indicated good platform performance both in terms of withinsample repeatability and between-lines reproducibility. Validation on publicly available NCI60 dataset showed similar patterns of variability, suggesting good reproducibility between experiments. Future research will explore the possibility to use this method to compare normalization methods as well as genomic platforms. © Springer-Verlag 2012.