Abstract：Co-expression analysis is an important approach to explore the genes responsible for different traits in large biological data set. In this study, we calculated the expression abundance of genes in 10 different tissues of cucumber (Cucumis sativus L.) using RNA-Seq data and removed the genes whose maximal fragments per kilobase of exon per million fragments mapped (FPKM) values were less than 5 across 10 different tissues. Then, co-expression modules were detected according to the correlation and TO (Topological Overlap) value between genes by applying WGCNA package in R project. Finally, 1 134 modules were obtained including 16 924 genes, of which 839 modules were selected for their mean correlation coefficients more than 0.9, getting 11 844 genes in total. The great functional and morphological variation in plant tissue types arises from the differential regulation of a finite set of genomic transcripts. The relationship analysis between modules and tissue types found 323 tissue-correlated modules including 5 784 genes. These modules highly correlated with tissues (r＞0.65). Using the topGO package in R project, we identified Gene Ontology (GO) terms that appeared in modules more frequently than expected. Nine of the 10 tissues had correlated-modules highly enriched in GO biological processes respectively (Fisher's exact test, P＜0.000 1), except tendril. GO enrichment analysis (Fisher's exact test, P＜0.000 1) of tissue-specific modules showed some specific genes were related to different tissues. The overrepresented GO biological processes in tissue-specific modules was often consistent with known tissue attributes. Positional clusters were also found in the modules with size ranging from 2 to 5 genes, most of which contained 2 genes. The physical distance between clustered genes in one module on the chromosome was less than 25 kb. The clustered genes often had similar structure or function. As an example, we got 3 modules in leaf and 1 module in stem correlated to the biosynthesis pathway of cucurbitacin, and 10 genes were published in previous research. These 4 modules were enriched in the biological process of terpenoid and isoprenoid which were the precursors of cucurbitacin. We also found that module M107 collected most of genes (7/10) associated with biosynthesis pathway of cucurbitacin which were in 2 clusters, which indicated that genes having similar function were often adjacently located on the chromosome. This study identified some important co-expression modules in cucumber through transcriptome and network analysis and provided a reference for similar studies in the future.
