RNA-Seq Evaluating Several Custom Microarrays Background Correction and Gene Expression Data Normalization Systems

Microarray gene expression technologies represents a widely used tool in transcriptomics and genomics studies worldwide. Even if this technology exhibits a low dynamic range as well as a feeble sensitivity and specificity (limited performances) with respect to RNA sequencing (RNA-seq) methodology in whole transcriptomic and/or genomic studies; it is noteworthy to underline the stability of the former (microarrays) because of their well-established biostatistics and bioinformatics analysis schemes. Several studies shown that inadequate data pre-processing as regards microarray gene expression data analysis; i.e. inadequate gene expression data normalization (DN) and scarce noise background subtraction (BS), might compromise microarray aptitude in calling correctly significantly differentially expressed genes (DEGs). Here, we were interested in assessing the performance of 20 different microarrays background correction and gene expression data normalisation arrangements from R software “linear models for microarray and RNA-seq data analysis” package, by comparing the number of differentially expressed genes detected by our previous developed custom microarray designs and RNA-seq platform. The present study basing exclusively on several clustering and principal component analysis (PCA) as well as descriptive and inferential statistic surveys, developed in the R programing environment, suggested a predominance of microarray data normalisation systems with respect to noise background correction procedure. Although, all processed background subtraction and gene expression data normalization arrangement (BS+DN) claimed to improve the agreement (sensitivity) between microarrays and RNA-seq in calling DEGs; quantile normalisation procedure applied to our processed custom microarray designs has been recorded as exhibiting the best sensitivity (p-value<0.05), since discriminates the highest number of DEGs in agreement with RNA-seq as opposed to the others analysed microarray gene expression data normalisation systems. In conclusion our findings confirmed the pre-eminence of data pre-processing procedure in microarray gene expression profiling analysis according a priority to data normalisation procedure and suggested the stability of quantile normalisation system with respect to the others processed normalisation arrangements in the present executed gene expression comparative study.