Drug-Path: a database for drug-induced pathways

Hui Zeng, Chengxiang Qiu, Qinghua CuiDatabase 2015 · DOI

TL;DR

This paper introduces Drug-Path, a novel database that catalogs pathways affected by drugs, going beyond traditional drug-action or metabolism pathways. By analyzing gene expression data, Drug-Path provides a valuable resource for understanding drug mechanisms and facilitating drug repurposing.

What Problem Does This Paper Address?

Understanding how drugs interact with biological systems is crucial for developing new therapies and finding new uses for existing drugs (drug repurposing). While existing databases focus on direct drug-target interactions or metabolic pathways, they often miss the broader cellular responses. Recent advances in high-throughput technologies like microarrays and RNA-sequencing have generated vast amounts of data showing how drugs alter gene expression. These changes in gene expression can activate or repress entire pathways, offering a more comprehensive view of a drug's impact. However, a centralized resource for these "drug-induced pathways" was lacking, hindering research in this area.

Methods

The authors developed Drug-Path by leveraging the Connectivity Map (CMap) database, which contains gene expression profiles for over 1000 drugs. Their approach involved:

  1. Identifying Deregulated Genes: For each drug instance in CMap, they identified genes that were significantly upregulated (fold-change > 2) or downregulated (fold-change < 0.5) compared to a control. These lists were combined for each drug.
  2. Pathway Enrichment Analysis: Using Fisher's exact test, they performed enrichment analysis on the upregulated and downregulated gene sets against 94 human KEGG pathways. This identified which pathways were statistically overrepresented by the drug-induced genes.
  3. Database Construction: The identified drug-induced pathways (both upregulated and downregulated) were stored in a SQLite database.
  4. Web Interface Development: A user-friendly web interface was built using Django (a Python web framework) to allow users to easily search, visualize, and download the data.

Datasets

  • Connectivity Map (CMap): This database provided the drug-induced gene expression profiles for over 1000 drugs. Specifically, they used data from 1294 drugs that had undergone transcriptional microarray experiments on MCF7 cell lines.
  • KEGG Pathways: 94 human KEGG pathways were used as the reference for the pathway enrichment analysis.

Key Insights and Ideas

  • Beyond Direct Action: Drugs don't just interact with their direct targets; they can trigger complex cascades of gene expression changes that affect entire cellular pathways. Drug-Path captures these broader, indirect effects.
  • Upregulated vs. Downregulated Pathways: A single drug can induce both activation (upregulation) and repression (downregulation) of different pathways, providing a more nuanced understanding of its cellular impact.
  • Potential for Drug Repurposing: By identifying pathways affected by drugs, researchers can hypothesize new therapeutic uses for existing drugs based on their known pathway modulations.

Key Results

  • Drug-Path contains 243,272 drug-pathway pairs, linking 1294 drugs to 94 KEGG pathways.
  • Each drug is associated with two categories of induced pathways: upregulated and downregulated.
  • The drug daunorubicin impacted the largest number of pathways, affecting up to 40.
  • The pathway "olfactory transduction" was the most frequently induced pathway, affected by over 1000 drugs.
  • The database allows users to visualize pathways with highlighted deregulated genes (yellow for upregulated, red for downregulated).

Limitations and Future Work

  • Limited Drug Coverage: The current database includes data for 1294 drugs, which is only a fraction of all available drugs. The authors suggest integrating data from other sources like GEO to expand coverage.
  • KEGG Pathway Focus: The current analysis is limited to KEGG pathways. Future work could include integrating other pathway annotations like BioCarta and Gene Ontology to provide a more comprehensive view.
  • Future Improvements: The authors plan to enhance Drug-Path by incorporating more gene annotations and potentially expanding the drug coverage. They believe Drug-Path will serve as a valuable resource for understanding drug mechanisms.

Summary updated

Abstract

Some databases for drug-associated pathways have been built and are publicly available. However, the pathways curated in most of these databases are drug-action or drug-metabolism pathways. In recent years, high-throughput technologies such as microarray and RNA-sequencing have produced lots of drug-induced gene expression profiles. Interestingly, drug-induced gene expression profile frequently show distinct patterns, indicating that drugs normally induce the activation or repression of distinct pathways. Therefore, these pathways contribute to study the mechanisms of drugs and drug-repurposing. Here, we present Drug-Path, a database of drug-induced pathways, which was generated by KEGG pathway enrichment analysis for drug-induced upregulated genes and downregulated genes based on drug-induced gene expression datasets in Connectivity Map. Drug-Path provides user-friendly interfaces to retrieve, visualize and download the drug-induced pathway data in the database. In addition, the genes deregulated by a given drug are highlighted in the pathways. All data were organized using SQLite. The web site was implemented using Django, a Python web framework. Finally, we believe that this database will be useful for related researches.
BibTeX
@inproceedings{zeng2015drug,
  author = {Hui Zeng and Chengxiang Qiu and Qinghua Cui},
  title = {{Drug-Path: a database for drug-induced pathways}},
  booktitle = {Database},
  year = {2015},
  doi = {10.1093/database/bav061},
}
Figure