Systems Pharmacology-Based Approach of Connecting Disease Genes in Genome-Wide Association Studies with Traditional Chinese Medicine

Jihye Kim, Minjae Yoo, Jimin Shin, Hyunmin Kim, Jaewoo Kang, Aik Choon TanInternational Journal of Genomics 2018 · DOI

TL;DR

This paper introduces a systems pharmacology approach to bridge the gap between genes identified in genome-wide association studies (GWAS) for various diseases and the molecular targets of Traditional Chinese Medicine (TCM). By building integrated networks, they identified 115 promising disease-TCM-gene pairs, suggesting novel drug repurposing opportunities for TCM, and developed a web application to make these findings accessible.

What Problem Does This Paper Address?

Modern medicine often focuses on single-target drugs, making their mechanisms of action (MoA) relatively straightforward to study. In contrast, Traditional Chinese Medicine (TCM), practiced for millennia, is known for its "multicomponent, multitarget" nature, making its MoA complex and difficult to elucidate. This complexity, however, also presents an opportunity for drug repurposing. Simultaneously, Genome-Wide Association Studies (GWAS) have identified numerous genes associated with various diseases, but translating these genetic findings into clinical applications remains a challenge. This paper aims to connect these two areas by using systems pharmacology to link disease-associated genes from GWAS with the known targets of TCM components, thereby uncovering potential new uses for TCM.

Methods

The researchers employed a systems pharmacology framework, integrating multiple data sources and computational analyses:

  • TCM Target Gene Identification: They analyzed microarray gene expression data from 102 TCM components treated on MCF7 breast cancer cells. Differentially expressed genes were identified to define the target genes for each TCM component.
  • GWAS Disease-Gene Network Construction: They utilized the NHGRI-EBI GWAS Catalog, which contains data from 2558 studies, to build a network of genes associated with various diseases. To standardize disease terms, they mapped GWAS disease names to Medical Subject Headings (MeSH) terms, focusing on the "Disease [C]" category.
  • Protein-Protein Interaction (PPI) Expansion: To enrich the GWAS disease-gene network, they used the STRING database to retrieve additional genes with high-confidence interactions (score > 0.9) for the initial disease-associated genes.
  • Statistical Framework for Prioritization: A permutation test was used to statistically assess the significance of a TCM's potential to treat a specific disease. This involved calculating a nominal p-value (P) and then applying Benjamini-Hochberg correction to obtain adjusted p-values (q-values). A Z-score was also calculated to compare TCM-target genes against GWAS disease genes.
  • Web Application Development: The identified connections were implemented in a user-friendly web application called "TCM-Disease," part of the larger "TCM hub."
  • Network Visualization: Cytoscape was used to visualize the constructed networks.
  • Connectivity Map Analysis: The Connectivity Map (CMap) was queried to find potential MoA evidence for TCM components against GWAS diseases.

Datasets

  • TCM Gene Expression Data: Microarray data from 102 TCM component-induced gene expression studies (GSE85871) from the Gene Expression Omnibus (GEO).
  • GWAS Data: The NHGRI-EBI GWAS Catalog (data version v1.0, associations_e89_r2017-06-19.tsv), containing data from 2558 studies and 1987 diseases/traits.
  • Protein-Protein Interaction Data: STRING database.
  • Medical Subject Headings (MeSH): Used for disease term mapping.
  • Connectivity Map (CMap): Used for MoA analysis.

Key Insights and Ideas

  • Bridging Ancient Wisdom and Modern Genomics: The core idea is to leverage modern high-throughput omics data (GWAS, gene expression) to scientifically validate and expand upon the empirical knowledge embedded in TCM.
  • Polypharmacology as a Feature, Not a Bug: The inherent complexity of TCM (multiple compounds, multiple targets) is reframed as a strength, making it a rich source for drug repurposing, especially for complex diseases.
  • Computational Integration for Discovery: The paper demonstrates the power of integrating diverse biological datasets (gene expression, GWAS, PPIs) through a systems pharmacology lens to uncover non-obvious connections.
  • Prioritization is Key: Simply identifying targets isn't enough; the statistical framework (permutation testing, q-values, Z-scores) is crucial for prioritizing the most likely and significant disease-TCM associations.
  • Web-Based Accessibility: Creating a dedicated web application (TCM-Disease) democratizes access to these complex computational findings, enabling researchers to explore potential TCM-based therapies.

Key Results

  • The study constructed a TCM-target gene network with 102 TCM components and 7380 target genes, and a GWAS disease-gene network with 293 diseases and 7417 disease genes.
  • They analyzed 14,713 GWAS disease-TCM-target gene pairs.
  • 115 GWAS disease-TCM-target gene pairs were identified with a q-value < 0.2.
  • These pairs involved 40 TCMs and 42 GWAS diseases.
  • Several identified pairs were validated with existing literature evidence, including:
  • Japonicone A for lymphoma treatment.
  • Artemisinin for type 1 diabetes mellitus.
  • Sanguinarine for breast cancer treatment.
  • Berberine hydrochloride for preeclampsia.
  • Tanshinone IIA for obesity treatment.
  • The TCM-Disease web application was developed and made publicly available.

Limitations and Future Work

  • Limited TCM Coverage: The study analyzed only 102 TCM components, which may limit the comprehensiveness of the TCM-target gene networks.
  • Limited Disease Coverage: The investigation was restricted to 293 GWAS diseases, potentially limiting the scope of disease-gene networks.
  • Validation Basis: Current validation relies on literature evidence and experimental results.

Future Work:

  • Expand the number of TCM components by mining additional GEO data and established drug-target databases (TCMID, TCM-MeSH, PharmDB-K, DSigDB).
  • Integrate manually curated targets to enhance TCM-target gene networks.
  • Expand disease-gene networks by using data from DisGeNet, which contains a much larger number of gene-disease associations. This is expected to increase the coverage of diseases and potentially reveal more novel disease-TCM genes.

Summary updated

Abstract

Traditional Chinese medicine (TCM) originated in ancient China has been practiced over thousands of years for treating various symptoms and diseases. However, the molecular mechanisms of TCM in treating these diseases remain unknown. In this study, we employ a systems pharmacology-based approach for connecting GWAS diseases with TCM for potential drug repurposing and repositioning. We studied 102 TCM components and their target genes by analyzing microarray gene expression experiments. We constructed disease-gene networks from 2558 GWAS studies. We applied a systems pharmacology approach to prioritize disease-target genes. Using this bioinformatics approach, we analyzed 14,713 GWAS disease-TCM-target gene pairs and identified 115 disease-gene pairs with <i>q</i> value < 0.2. We validated several of these GWAS disease-TCM-target gene pairs with literature evidence, demonstrating that this computational approach could reveal novel indications for TCM. We also develop TCM-Disease web application to facilitate the traditional Chinese medicine drug repurposing efforts. Systems pharmacology is a promising approach for connecting GWAS diseases with TCM for potential drug repurposing and repositioning. The computational approaches described in this study could be easily expandable to other disease-gene network analysis.
BibTeX
@article{kim2018systems,
  author = {Jihye Kim and Minjae Yoo and Jimin Shin and Hyunmin Kim and Jaewoo Kang and Aik Choon Tan},
  title = {{Systems Pharmacology-Based Approach of Connecting Disease Genes in Genome-Wide Association Studies with Traditional Chinese Medicine}},
  journal = {International Journal of Genomics},
  year = {2018},
  doi = {10.1155/2018/7697356},
}
Figure