The Network Module-based Statistic Identification Of Gene Expression In RNA-Seq

With the rapid advances and lower cost in high-throughput sequencing,RNA-Seq has been a powerful technology in transcriptome study.Differential expression analysis is the basis for the downstream analysis of transcriptome,but the existing statistical methods are designed on the basis of single gene-level.Nevertheless,many molecular and clinical phenotypes are proved to be associated with a subtle but coordinated regulation in a biological network module rather than single gene.Recently the network module-based statistical methods have emerged as a helpful way to compare,integrate and interpret data of microarrays and proteomics,and exhibited a high power to identify the differential expression modules.Nevertheless,network-based differential expression analysis approach has not been addressed for RNA-Seq data.In this study,we proposed a network module-based generalized linear model for differential expression analysis with the count-based sequence data from RNA-Seq.Simulation study was conducted to prove that the proposed method can improve the power of module identification.The new model was also applied to human tissue and cancer samples to identify the tissue-specific modules and cancer-associated modules through differentially expression analysis.Besides,in comparison with the model on the single gene-level,our method can identify more robust profiles of significantly altered biological modules.Finally,a R package was designed and constructed to implement our module-based statistical models for RNASeq data as well as for microarray and shotgun proteomics data.In summary,our statistical model takes into account the interactions among genes,leading to the effective identification of subtle but coordinate changes on a system level.The model also provides a powerful framework for integrating the expression information in the network system and analyzing complex experimental designs for RNA-Seq count data,which woud give novel insight into the identification of potential differential expression modules,the pathogenesis researches of complex diseases and new drug target researches.