A network biology workflow to study transcriptomics data of the diabetic liver

Abstract

Background: Nowadays a broad collection of transcriptomics data is publicly available in online repositories. Methods for analyzing these data often aim at deciphering the influence of gene expression at the process level. Biological pathway diagrams depict known processes and capture the interactions of gene products and metabolites, information that is essential for the computational analysis and interpretation of transcriptomics data.The present study describes a comprehensive network biology workflow that integrates differential gene expression in the human diabetic liver with pathway information by building a network of interconnected pathways. Worldwide, the incidence of type 2 diabetes mellitus is increasing dramatically, and to better understand this multifactorial disease, more insight into the concerted action of the disease-related processes is needed. The liver is a key player in metabolic diseases and diabetic patients often develop non-alcoholic fatty liver disease.

Results: A publicly available dataset comparing the liver transcriptome from lean and healthy vs. obese and insulin-resistant subjects was selected after a thorough analysis. Pathway analysis revealed seven significantly altered pathways in the WikiPathways human pathway collection. These pathways were then merged into one combined network with 408 gene products, 38 metabolites and 5 pathway nodes. Further analysis highlighted 17 nodes present in multiple pathways, and revealed the connections between different pathways in the network. The integration of transcription factor-gene interactions from the ENCODE project identified new links between the pathways on a regulatory level. The extension of the network with known drug-target interactions from DrugBank allows for a more complete study of drug actions and helps with the identification of other drugs that target proteins up- or downstream which might interfere with the action or efficiency of a drug.

Conclusions: The described network biology workflow uses state-of-the-art pathway and network analysis methods to study the rewiring of the diabetic liver. The integration of experimental data and knowledge on disease-affected biological pathways, including regulatory elements like transcription factors or drugs, leads to improved insights and a clearer illustration of the overall process. It also provides a resource for building new hypotheses for further follow-up studies. The approach is highly generic and can be applied in different research fields.

Figure 1

Transcriptomics network biology workflow. The described workflow uses four different tools that can be easily connected by taking the output of one tool and importing it in the next tool. No intermediate steps are required.

Figure 2

Visualization of two pathways relevant for drug treatment of T2DM. Gene expression is visualized on (A) AMPK Signaling pathway, http://www.wikipathways.org/instance/WP1403 and (B) Statin pathway, http://www.wikipathways.org/instance/WP430 from WikiPathways. The visualization of the gene product boxes in the pathways is split into two parts, (1) the log2 FC in the left part of the box (blue is down-regulated over white is not changed to red is up-regulated) and (2) the p-value in the right part of the box (green when significant). Pathway elements including metabolites that have not been measured in the selected dataset are gray.

Figure 3

Nodes linking the seven significantly changed pathways. Each pathway is represented as a yellow rounded rectangle. Gene products and metabolites are visualized as ellipses and octagons, respectively. The transcription dataset is visualized on the gene nodes in the network using a color gradient from blue (down-regulated) over white (not changed) to red (up-regulated). Nodes with a significant p-value (< 0.05) have a light-green border color. Most nodes linking multiple pathways are either up-regulated (e.g. SREBF1, CDKN1A, AGPAT2) or not altered significantly (e.g. LPL, HMGCR).

Figure 4

TF regulation in the diabetic fatty liver pathways. Using CyTargetLinker, sixteen TFs have been identified in the seven pathways. TFs are visualized as rounded rectangles and their target genes as circles colored based on their presence in different pathways. 56 genes are targeted by only one TF and 33 genes are targeted by 2 or more TFs. Light-blue edges indicate regulation of TFs by other TFs.

Figure 5

Significant amount of antidiabetic drugs targeting gene products in the network. The network has been extended with drug-target interactions from DrugBank 4 using CyTargetLinker. Nodes present in only one pathway and not targeted by any drugs have been grouped in pathway nodes (yellow rounded rectangles). Drugs targeting genes in the network are indicated as blue rectangles, drugs associated with diabetes are colored in red, micronutrients/dietary supplements in green, drugs related to immune response in orange and anticholesteremic agents in purple. Diabetes related drugs target 7 gene products in the network: INSR (8 drugs), PPARG (5 drugs), RB1 (2 drugs), ABCA1 (1 drug), CPT1A (1 drug), PRKAA1 (1 drug), PRKAB1 (1 drug).