Gene Expression Data Analysis Across Multiple Experimental Platforms

In the transcriptomics study,gene expression analysis is one of the most basic research approaches.Expression calculation and differential expression(DE)analysis of gene and isoform provide possibilities to understand gene and isoform functions and regulating mechanism.As the two mainstream large-scale technologies for gene expression measurement,microarray and RNA-Seq baesd on high-throughput sequencing technology,are widely used in transcriptomics study.Large amounts of expression data have been generated from the technologies,and this makes integrating expression data from multiple platforms become possible.The main work of this paper includes the following two parts:(1)Comparison of gene and isoform expression analysis across multiple platforms.We first introduced the technical principles of the four mainstream experimental platforms: Affymetrix's traditional 3' GeneChip,Exon array,Human Transcriptome Array 2.0 and Illumina platform based on RNA-Seq.We then reviewed the mainstream analysis methods on each platform for the calculation of gene expression levels and DE analysis.We also used a well-defined benchmark data set to compare the expression measurement and DE analysis across various platforms.The comparison results provide a reference for researchers to choose experimental platforms and data analysis methods.(2)Differential expression analysis based on integrating transcriptome expression data from multiple platforms.Considering the problems of the existing methods for integrating expression data from multiple platforms,we propose a new model,mpDE(multi-platform Differential Expression model),for DE detection by integrating expression data from multiple platforms.The mpDE method integrates the expression data and the associated measurement error from different platforms and considers the variability of biological replicates or technical replicates under different conditions for the same platform to improve the accuracy of DE detection.We applied mpDE to three human datasets and compared its performance with each single platform and other multi-platform analysis methods.The results show that mpDE is able to obtain more accurate and sensitive DE analysis results.