Colorectal cancer drug target prediction using ontology-based inference and network analysis

Cui Tao, Jingchun Sun, W. Jim Zheng, Junjie Chen, Hua XuDatabase 2015 · DOI

TL;DR

This paper introduces a novel computational framework that combines semantic web ontologies with protein-protein interaction (PPI) network analysis to predict new drug targets for colorectal cancer (CRC). By integrating diverse biological data, their approach successfully identified 15 promising drug targets, including known targets and novel candidates supported by literature.

What Problem Does This Paper Address?

Developing new drugs is a notoriously slow and expensive process. Identifying effective drug targets is a crucial bottleneck. While many data sources exist (genomic, pharmaceutical, pathway information), integrating these disparate datasets to uncover hidden relationships and predict novel targets is challenging. Traditional drug discovery methods are often inefficient, and even drug repurposing (finding new uses for existing drugs) can be limited by the complexity of biological mechanisms. This paper aims to address this by creating a more systematic and integrative approach to drug target prediction, using colorectal cancer as a case study.

Methods

The authors developed a two-pronged computational framework:

  1. Ontology-based Inference: They built a Web Ontology Language (OWL) ontology to represent relationships between drugs, diseases, genes, pathways, and Single Nucleotide Polymorphisms (SNPs). This ontology was populated with data from the PharmGKB database. Using Description Logic (DL) rules, they inferred potential drug targets by identifying genes that are closely associated with known CRC drugs and CRC-related entities within this structured knowledge base.
  2. Network-Assisted Gene Ranking: To prioritize the inferred potential targets, they employed a network-based approach. They mapped the potential drug targets and known CRC disease genes onto a human protein-protein interaction (PPI) network. Candidate targets were then ranked based on the "neighborhood" of CRC disease genes surrounding them in the PPI network, hypothesizing that targets closer to causal genes are more likely to be effective. A robust rank aggregation (RRA) method was used to combine rankings from different network proximity levels (first, second, and third-degree neighbors).

Datasets

  • PharmGKB: Used to construct the initial ontology, providing data on drugs, diseases, genes, pathways, and SNPs.
  • FDA-approved CRC drugs: Collected from the National Cancer Institute's cancer drug list and verified with FDA label files.
  • DrugBank and Therapeutic Target Database (TTD): Used to extract known drug targets for FDA-approved drugs.
  • KEGG (Kyoto Encyclopedia of Genes and Genomes): Used for drug classification (ATC) and for pathway enrichment analysis.
  • Cancer Gene Census (CGC), Online Mendelian Inheritance in Man (OMIM), and Genetic Association Database (GAD): Used to compile a list of CRC disease genes.
  • Human Protein-Protein Interaction (PPI) Network: Obtained from the Protein Interaction Network Analysis platform (PINA v2.0).

Key Insights and Ideas

  • The Power of Integration: The core idea is that by semantically linking diverse biological data (drugs, genes, diseases, pathways, SNPs) and then analyzing their network proximity, we can uncover more robust drug target candidates than by looking at individual data types alone.
  • Ontology as a Knowledge Graph: Using OWL and DL rules allows for a structured, machine-understandable representation of complex biological relationships, enabling automated inference of potential targets.
  • Network Proximity as a Prioritization Metric: The hypothesis that drug targets "closer" to known disease genes in a PPI network are more likely to be effective is a powerful heuristic for filtering and ranking candidates.
  • CRC as a Valid Use Case: Demonstrating the framework's efficacy on a well-studied cancer like CRC provides strong validation for its potential applicability to other complex diseases.

Key Results

  • The ontology-based inference identified 113 potential CRC drug targets.
  • These 113 genes showed significant enrichment with known CRC disease genes and known CRC drug targets.
  • Pathway enrichment analysis revealed that these genes were significantly involved in pathways related to drug metabolism and cancer, as well as carbohydrate metabolism, aligning with existing research.
  • Network-assisted ranking prioritized 15 promising drug targets from the initial 113.
  • Among these 15, EGFR, TOP1, and VEGFA are already known targets of FDA-approved CRC drugs.
  • CCND1 and PTGS2 were identified as promising candidates supported by literature for their relevance to CRC or as potential drug targets.

Limitations and Future Work

  • Data Source Expansion: The authors suggest integrating more data sources (e.g., DrugBank, KEGG) into the ontology to provide more inference options.
  • Relationship Granularity: The current system treats all relationship types equally. Future work could differentiate between relationship types (e.g., direct vs. indirect, causal vs. correlational) for more nuanced inference.
  • Disease Gene and Drug Target Completeness: The accuracy and comprehensiveness of the input disease gene and drug target lists are critical for network-based prioritization. The authors plan to expand these lists by manually checking publications and FDA labels.
  • Association Types: Future research will explore whether disease genes and targets are preferentially found in specific types of associations within PPI networks (e.g., physical vs. enzymatic).
  • Broader Applications: The framework is intended to be applicable to other diseases beyond CRC, and future work includes expanding predictions to drug repurposing and potentially pursuing clinical trials for promising candidates.

Summary updated

Abstract

Identification of novel drug targets is a critical step in drug development. Many recent studies have produced multiple types of data, which provides an opportunity to mine the relationships among them to predict drug targets. In this study, we present a novel integrative approach that combines ontology reasoning with network-assisted gene ranking to predict new drug targets. We utilized colorectal cancer (CRC) as a proof-of-concept use case to illustrate the approach. Starting from FDA-approved CRC drugs and the relationships among disease, drug, gene, pathway, and SNP in an ontology representing PharmGKB data, we inferred 113 potential CRC drug targets. We further prioritized these genes based on their relationships with CRC disease genes in the context of human protein-protein interaction networks. Thus, among the 113 potential drug targets, 15 were selected as the promising drug targets, including some genes that are supported by previous studies. Among them, EGFR, TOP1 and VEGFA are known targets of FDA-approved drugs. Additionally, CCND1 (cyclin D1), and PTGS2 (prostaglandin-endoperoxide synthase 2) have reported to be relevant to CRC or as potential drug targets based on the literature search. These results indicate that our approach is promising for drug target prediction for CRC treatment, which might be useful for other cancer therapeutics.
BibTeX
@inproceedings{tao2015colorectal,
  author = {Cui Tao and Jingchun Sun and W. Jim Zheng and Junjie Chen and Hua Xu},
  title = {{Colorectal cancer drug target prediction using ontology-based inference and network analysis}},
  booktitle = {Database},
  year = {2015},
  doi = {10.1093/database/bav015},
}
Figure