Bioinformatic Analysis Of Dynamic Transcriptome And The Construction Of Expression Regulation Network Multi-step Colorectal Cancer

[Background]Colorectal cancer (CRC) is a common malignant gastroenterological cancer. The molecular mechanism of CRC tumorigenesis has been one of the most important fields in cancer research. The CRC carcinogenesis is a complicated process with poly-genic factors. Its malignant phenotypes expressed in quantitive changes are considered to be related to the accumulation of genetic and environmental alteration. The multi-step and poly-genic characteristics of CRC gradually make it the "tumor model " for fundamental cancer research.Transcriptome is a set of all RNA transcripts including mRNA, tRNA, rRNA and non-coding RNA such as miRNA produced in one or a population of certain type of cell. Unlike genome, which is roughly fixed for a certain type of cell, the transcriptome can vary with external environmental condition and it is considered to be highly dynamic. When the cell suffers different physiologic or pathologic stimuli, its transcriptome will change dramatically. Transcriptomic changes inherit from genomic information and take place before the proteomic level. Understanding of this crucial stage of genomic information process is of most importance for us to reveal the mechanism of life phenomena including tumorigenesis. In our study, we purposed the idea of dynamic transcriptome and put it forward to the study of CRC transcriptomics and establishment of gene expression regulation network.Gene microarray is one the most important tool for transcriptomic study which could simultaneously detect thousands of genes expression. The high-throughput technologies combined with bioinformatics approach would be the best way for screening of the potential differentially expressed genes. The online database with massive gene functional and pathway information will support the further analysis and provide valuable in-silico advice for further research such as biomarker selection and prognosis prediction. Based upon the theory above, in our study, the whole-genome olig-nucleotide gene microarray and miRNA expression microarray were applied to examine the expression of multi-step colorectal cancer and adjacent normal mucosa specimen consisted of 4 TNM stage (with each stage 5 replicates). Further bioinformatics analysis was carried out based on theory of network biology for thoroughly data mining and data integration.[Analysis of differentially expressed genes between tumor and its adjacent normal tissue of different stage]TwoClassDif (RVM-T test) method is applied to analyze the differentially expressed genes in 4 different TNM stage of CRC. With the cutoff of P-value<0.05 and FDR<0.05, we obtained 4 groups of data of significant differentially expressed genes:698,547,609 and 713genes were upregulated whereas 104,235,143 and 158 genes were downregulated in stageⅠ,Ⅱ,ⅢandⅣrespectively.GO analysis showed 82,90,106 and 58 significantly upregulated GO, and 45,65,45and 84 downregulated GO in 4 stages respectively. Pathway analysis showed 7,9,19 and 17 significant pathways of upregulated genes, and 19,32,16 and 26 significant pathway of downregulated genes in 4 stages, respectively (P-value<0.05 and FDR<0.05). The PPAR pathway in stageⅠ, Focal adhesion pathway in stageⅡ, Cytokine-cytokine receptor interaction pathway in stageⅢand Wnt signaling pathway in stageⅣwere significantly altered.According to the graph theory and the relationship provided by KEGG pathway database, we built the path-net showing the interconnection of the pathways of the 4 stages. The main pathways affected in stageⅠwere MAPK signaling pathway, Cytokine-cytokine receptor interaction. The main pathway affected in stageⅡwere citrate cycle, valine leucine and isoleusine degradation. The main pathways affected in StageⅢwere cytokine-cytokine receptor interaction and apoptosis. The main pathways affected in StageⅣwere MAPK signaling pathway, calcium signaling pathway. It is noticed that the MAPK signaling pathway, cytokine-cytokine receptor interaction and apoptosis pathway were of most importance to the progression of CRC.Gene signal transduction network, the signal-net were established based on KEGG database about the interaction between different genes product and theory of network biology. The signal-net referred to the inter-genes signal communication between the differentially expressed genes. The network could provide us with the main effect genes such as ITGB1, HASPA9B, PAG and more. These genes would play important role in different stage of CRC carcinogenesis.[Analysis of multi-Class differentially expressed genes of different stage]The gene expression data of normal, stageⅠ,Ⅱ,ⅢandⅣwere set up as five time point according to the time-series of the CRC progression. MultiClassDif method (RVM-F test) is applied to screen the dynamic differentially expressed genes.2858 multi-class differentially expressed genes were obtained with the cutoff of p<0.05 and FDR＜0.05.Serial Test Cluster analysis (STC) is applied to analyze the dynamic gene expression pattern of the multi-class differentially expressed genes. 20 out of 62 patterns were identified as significant expression pattern (p<0.05/80). Among them, pattern No.67, No.41 and No.80 are important to biological understanding.Furthermore, the genes from the 20 significant expression patterns were analyzed according to the similarity of gene expression. "Cluster coefficient" and "betweenness centrality" are the parameters that judge the the gene's ability of interconnection with others. We build the Dynamic-Gene-net according to these parameters and with calculation we obtained some of the most important genes in the network such as AFURS1, E2F5, C14orf104 and more.[Analysis of Multi-class differentially expressed miRNA of different stages]Similarly we choose the normal, stageⅠ,Ⅱ,Ⅲ,Ⅳas the 5 time point to screen the differentially expressed miRNAs with MultiClassDif method (RVM-F test).55 miRNAs were obtained (p<0.05 and FDR<0.05) and 42 had their records in Targetscan database.STC method was applied to analyze the dynamic expression pattern of miRNAs as well.2 out of 22 patterns were significant (p<0.05/80):the pattern No.18 and No.59.[Integration analysis of miRNA and mRNA of different CRC stages]The Targetscan database of miRNA target gene prediction was applied for the 42 miRNAs target prediction and there were 5930 miRNAs'target genes on total. Intersection of 5930 target genes and 2858 multi-class differentially expressed genes were calculated with 605 genes. These genes were further carried out with GO and pathway analysis.51 significant GO referred to 307 genes were obtained with GO analysis (p<0.01 and FDR<0.05). The miRNAs and GO were connected to build the miRNA-GO-network which showed the target genes functional network according to their relationship to the miRNA. The differentially expressed miRNA and their target genes with significant GO were connected to form the miRNA-Gene-network which showed the relationship of miRNA and their related target genes. Two network concluded that the mir-524-5p, mir-429, mir-340, mir-124 are crucial to the network.Pathway analysis of 605 differentially expressed target genes of miRNAs showed 43 significant pathway referred to 167 genes (p<0.05 and FDR<0.05). Path-net of these significant pathway were build and showed the main effect pathways are MAPK signal pathway, Focal adhesion, adheren junction, Wnt signaling and pathway in cancer. These pathways are reported to be highly related to carcinogenesis.Based on the negative regulation relationship between the miRNA and its target genes, the negative correlation analysis was carried out with 2 significant miRNA expression pattern and 20 significant gene expression patterns. The miRNA-Gene-network were also build and the analysis result showed the expression patterns of mir-429, mir-490-3p, mir-18a and mir-18b were negative correlated with target genes ONECUT2, KCMF1, TRIM2 and so on. Real-time PCR validated the mir-18 negatively regulated the MDGA1 and NAV1 in transcriptional level. These result indicated the potential negatively regulating relationship between these miRNA and their target genes. [Conclusion]Investigation of the poly-genic factors and altereation of multiple pathways involved in the multi-step CRC progression can systematically uncover the dynamic mechanisim of tumorigenesis. The non-coding miRNAs play important role in posttranscriptional regulation during cancer initiation and promotion. We performed the gene and miRNA expression microarray experiment to analyze the two-class difference between cancer and its adjacent normal tissue as well as the multi-class difference of different time point of tumor progression. Bioinformatics with network biology approach comprehensively depict the miRNA-target mRNA regulation network in trancriptomic level. These results are important for us to understand the dynamic transcriptome of CRC and will guide us a way to further investigation of CRC carcinogenesis.