Performance assessment of different microarray designs using RNA-Seq as reference

Abstract In the past decade, the completion of sequencing of higher organisms has led to the development of whole transcriptome analysis techniques. Among the most important innovations in this field is the microarray technology. It allows to quantify the expression for thousand of genes simultaneously by measuring the hybridization from a tissue or cell of interest to probes immobilized on a solid surface. This powerful technology has applications in addressing many biological questions at genomic scale that were not approachable previously; however, the enormous size of microarray data sets leads to issues of experimental design and statistical analysis that are unfamiliar to many molecular biologists. The type of array used, the design of the biological experiment, the number of experimental replicates, and the statistical method for data analysis should all be chosen based on the scientific goals of the investigator. Here we compare two different strategies of array design (single replicate probe per transcript and multiple probes per transcript) in two highly customizable microarray platforms (CombiMatrix and NimbleGen). In this work we implemented a statistical methodology based on comparison of microarrays with RNA-Seq data to highlight the differences among different platforms and array designs and the causes of such differences. Our work showed that, the four analyzed microarray designs exhibited different advantages depending on the considered parameter (sensibility, specificity, accuracy and predictive positive values ).Thus, our results provide insights and guidance that can be used by researchers for properly selecting the approach more suitable to their scientific goal.