Interactive Toxicogenomics: Gene set discovery, clustering and analysis in Toxygates

Abstract

Toxygates was originally released as a user-friendly interface to enhance the accessibility of the large-scale toxicogenomics database, Open TG-GATEs, generated by the Japanese Toxicogenomics Project. Since the original release, significant new functionality has been added to enable users to perform sophisticated computational analysis with only modest bioinformatics skills. The new features include an orthologous mode for data comparison among different species, interactive clustering and heatmap visualisation, enrichment analysis of gene sets, and user data uploading. In a case study, we use these new functions to study the hepatotoxicity of peroxisome proliferator-activated receptor alpha (PPARα) agonist WY-14643. Our findings suggest that WY-14643 caused hypertrophy in the bile duct by intracellular Ca²⁺ dysregulation, which resulted in the induction of genes in a non-canonical WNT/Ca²⁺ signalling pathway. With this new release of Toxygates, we provide a suite of tools that allow anyone to carry out in-depth analysis of toxicogenomics in Open TG-GATEs, and of any other dataset that is uploaded.

Figure 1

(A) Many functions in Toxygates now communicate with each other by modifying or transferring gene sets. Currently displayed genes can be filtered, sorted and tested for enrichment, and saved as a named set. They can also be clustered, and by setting a dendrogram cutoff, new gene sets can be generated, as well as enriched together. Gene sets are saved as local data in the web browser. They may be manually edited, for example by adding pathways and GO terms, and saved gene sets may be synchronised with TargetMine for further analysis and long term storage. (B) The My Data screen. Here, users can upload, adjust, and delete their own expression data as sets of samples (called batches). (C) Interactive display of heatmap with dendrogram. After selecting compounds and experimental conditions and filtering genes of interest, a clustering may be performed and displayed as a heatmap in a new pop-up window by selecting “Show heatmap” from “Tools” pull-down menu at the top of the main display. The parameters of the clustering may be changed by using the controls on the right side of the display. (D) When the currently defined sample groups contain more than one platform (i.e. more than one species), the orthologous mode is activated. This uses pre-computed groups of probes (based on amino acid sequence similarity) to display orthologous genes together as a single row. This allows for meaningful cross-species analysis, which is especially useful since Open TG-GATEs contains both rat and human data. The high-resolution image file is available at Scientific Reports online.

Figure 2

Heatmap of gene expression profiles obtained by treatment of WY-14643. The dendrogram was made without log-transformation of the axis, according to a hierarchical clustering result (method: ward.D2, distance: Pearson). (A) Gene expression profile of M4vs8_cluster1 (probe: n = 268). The x-axis represents experimental conditions (from left to right, (1–8) WY-14643 3 hr, 6 hr, 9 hr, 8 day, 15 day, 29 day, 24 hr and 4 day; (9–11) amlodipine 6 hr, 24 hr and 3 day. The y-axis represents probes in M4vs8_cluster1. (B) Gene expression profile of M4vs8_cluster2 (probe: n = 188). The x-axis represents experimental conditions (from left to right): (1–3) amlodipine 24 hr, 6 hr and 3 day; (4–11) WY-14643 4 day, 3 hr, 6 hr, 15 day, 29 day, 8 day, 24 hr and 9 hr. The y-axis represents probes in M4vs8_cluster2. (C) Gene expression profile of M4vs8_cluster3 (probe: n = 162). The x-axis represents experimental conditions (from left to right, 1–7) WY-14643 29 day, 15 day, 8 day, 3 hr, 6 hr, 24 hr and 9 hr; (8–10) amlodipine 24 hr, 3 day and 6 hr; 11) WY-14643 4 day). The y-axis represents probes in M4vs8_cluster3. (D) Analytical workflow for obtaining three M4vs8_clusters in (Fig. 2A–C). A gene set was selected by filtering DEGs between WY-14643/M dose/4 day and WY-14643/M dose/8 day by Welch’s t-test (cutoff: p-value = 0.01). As columns, Open TG-GATEs samples of WY-14643 (single dose (30 mg/kg bw): 3 hr, 6 hr, 9 hr, 24 hr, repeat dose (30 mg/kg bw): 4 day, 8 day, 15 day, 29 day) and external samples of amlodipine (uploaded to Toxygates by “user data upload function”, L dose (0.2 mg/kg), 6 hr or 24 hr, and M dose (19 mg/kg), 3 day) were used. The clusters were obtained by hierarchical clustering (dendrogram cutoff: 6). The high-resolution image file is available at Scientific Reports online.

Figure 3

Hypothetical model of the hepatotoxicity caused by WY-14643. Genes in M4vs8_cluster1 were upregulated predominantly in M24hr and M4. This results in elevated sensitivity of beta-catenin independent Wnt signaling and induction of Ca²⁺ pathway-related genes, followed by decreasing intracellular Ca²⁺ and Ca²⁺ release from ER in M4. Upregulation of genes in M4vs8_cluster3, which were also upregulated by amlodipine, implies the Ca²⁺ dynamics dysregulation in M4. Also, temporary upregulation of genes in M4vs8_cluster2 in M4 implies UPR response, which is a survival signalling from ER stress that can be caused by Ca²⁺ depletion at ER. By long-term treatment of WY-14643 (M8, M15, M29), upregulation of genes in M4vs8_cluster1 (enriched pathways: “beta-catenin independent Wnt signalling”) and downregulation of genes in M4vs8_cluster 2 (enriched IPCs: “Protein processing in ER”) were observed. Since beta-catenin dependent Wnt signaling directly regulates cell cycle and beta-catenin independent Wnt signalling (non-canonical Wnt signalling, which is represented as “nc-Wnt signal” in Fig. 3) antagonises beta-catenin dependent Wnt signalling^, , it is implied that WY-14643 disrupts proper cell cycle regulation by inducing genes in beta-catenin independent Wnt signaling, while causing organelle proliferation by persistently activating PPARα signalling. The high-resolution image file is available at Scientific Reports online.