Integrative analysis of transcriptional regulatory network and copy number variation in intrahepatic cholangiocarcinoma

Abstract

Background: Transcriptional regulatory network (TRN) is used to study conditional regulatory relationships between transcriptional factors and genes. However few studies have tried to integrate genomic variation information such as copy number variation (CNV) with TRN to find causal disturbances in a network. Intrahepatic cholangiocarcinoma (ICC) is the second most common hepatic carcinoma with high malignancy and poor prognosis. Research about ICC is relatively limited comparing to hepatocellular carcinoma, and there are no approved gene therapeutic targets yet.

Method: We first constructed TRN of ICC (ICC-TRN) using forward-and-reverse combined engineering method, and then integrated copy number variation information with ICC-TRN to select CNV-related modules and constructed CNV-ICC-TRN. We also integrated CNV-ICC-TRN with KEGG signaling pathways to investigate how CNV genes disturb signaling pathways. At last, unsupervised clustering method was applied to classify samples into distinct classes.

Result: We obtained CNV-ICC-TRN containing 33 modules which were enriched in ICC-related signaling pathways. Integrated analysis of the regulatory network and signaling pathways illustrated that CNV might interrupt signaling through locating on either genomic sites of nodes or regulators of nodes in a signaling pathway. In the end, expression profiles of nodes in CNV-ICC-TRN were used to cluster the ICC patients into two robust groups with distinct biological function features.

Conclusion: Our work represents a primary effort to construct TRN in ICC, also a primary effort to try to identify key transcriptional modules based on their involvement of genetic variations shown by gene copy number variations (CNV). This kind of approach may bring the traditional studies of TRN based only on expression data one step further to genetic disturbance. Such kind of approach can easily be extended to other disease samples with appropriate data.

Figure 1. Workflow of integrative analysis of TRN and CNV in ICC.

Figure 2. Overview of module subtype and size in CNV-ICC-TRN.

In both A and B figures, blue color represents CNV-gene-only enriched module, green color represents CNV-TF-only regulated module, red color represents both CNV-TF regulated and CNV-gene enriched module.

Figure 3. Biological functions tackled in modules of CNV-ICC-TRN.

X-axis represents signaling pathways and IDs, y-axis represents the number of modules enriching to each pathway. Complete information is shown in Table S3.

Figure 4. Integrative analysis of CNV-ICC-TRN and KEGG signaling pathways.

(A). Integrated network of CNV-ICC-TRN and KEGG signaling pathways. Triangle represents TF, circle represents gene and rectangle represents signaling pathway; red color means gene inside CNV region, green color means gene outside CNV region. (B) Integrative analysis in MAPK signaling pathway. (C) Integrative analysis in Wnt signaling pathway. (D) Integrative analysis in TGF-β signaling pathway. In figure B, C, D, red edges are from CNV-ICC-TRN, and off-white edges are from signaling pathways.

Figure 5. ICC subclasses.

Based on expression profiles of genes in CNV-ICC-TRN, the non-negative matrix factorization–based algorithm divided ICC samples to two classes cluster I (right branch) and cluster P (left branch). This figure is heat map of differentially expressed genes between two classes.