Systems chemical biology and the Semantic Web: what they mean for the future of drug discovery research

David Wild, Ying Ding, Amit Sheth, Lee Harland, Eric Gifford, Michael S. LajinessDrug 2011 · DOI

TL;DR

This paper argues that integrating diverse biological and chemical data using Semantic Web technologies (RDF, OWL, SPARQL) is crucial for advancing systems chemical biology and revolutionizing drug discovery. By enabling sophisticated querying and analysis of interconnected data, these technologies can uncover complex drug-disease relationships and accelerate the identification of new therapeutics.

What Problem Does This Paper Address?

Traditional drug discovery, often relying on a "reductionist" approach of targeting single proteins, has faced challenges with drug failures and withdrawals. This highlights the need for a more holistic understanding of how small molecules interact with complex biological systems. The paper addresses the gap in effectively integrating and analyzing the vast and heterogeneous datasets generated in chemical biology and drug discovery. Existing methods struggle to connect disparate data sources, hindering the ability to uncover complex relationships and derive meaningful insights.

Methods

The paper champions the use of Semantic Web technologies as the solution for integrating and analyzing complex biological and chemical data. Key technologies discussed include:

  • RDF (Resource Description Framework): A standard for representing data as triples (subject-predicate-object), enabling flexible representation of relationships between entities like compounds, targets, and diseases.
  • OWL (Web Ontology Language): Used to define ontologies, which provide a shared vocabulary and formal representation of concepts and relationships within a domain, enabling richer data modeling and reasoning.
  • SPARQL: A query language for RDF, allowing for complex, integrative searches across multiple heterogeneous datasets, akin to SQL for relational databases but with enhanced capabilities.
  • Triple Stores: Databases optimized for storing and querying RDF data, enabling efficient management of large-scale semantic datasets.

The paper also discusses the application of graph theoretic algorithms and pathfinding algorithms on these semantically integrated networks to discover associations and infer relationships between entities, such as linking drugs to side effects through intermediate genes or targets.

Datasets

The paper references several publicly accessible databases that are relevant to chemical biology and drug discovery, including:

  • ChEMBL
  • ChemProt
  • PubChem
  • ChemSpider
  • UniProt
  • KEGG
  • DrugBank
  • PharmGKB
  • SIDER (for side-effect data)

The paper does not introduce new benchmark datasets but rather focuses on how existing datasets can be integrated and leveraged using Semantic Web technologies.

Key Insights and Ideas

  • The "Reductionist" Approach is Insufficient: The paper strongly advocates that understanding complex biological systems and drug interactions requires moving beyond single-target approaches.
  • Semantic Integration is Key: Simply aggregating data is not enough; data must be integrated semantically, meaning there's a shared understanding of the meaning and relationships between entities.
  • Semantic Web Technologies as Enablers: RDF, OWL, and SPARQL provide the foundational technologies for achieving this semantic integration, enabling sophisticated querying and knowledge discovery.
  • Networks Reveal Hidden Connections: Analyzing data as interconnected networks (e.g., drug-target-disease networks) can uncover non-obvious relationships and provide novel insights into drug mechanisms and side effects.
  • Bridging the Molecular and Clinical Spectra: The ultimate goal is to create a comprehensive "biomedical map" that connects molecular-level drug actions to clinical outcomes in humans.

Key Results

The paper focuses on the potential and methodology rather than presenting specific quantitative results from a single experiment. However, it highlights:

  • Demonstrated Successes: The paper points to successful applications of Semantic Web technologies in areas like predicting new drug targets, repurposing existing drugs, and identifying potential risk factors for drug side effects.
  • Example Visualization: Figure 2 illustrates a "constrained association search" linking the drug rosiglitazone to myocardial infarction through a network of genes and databases, demonstrating the power of pathfinding algorithms on semantically integrated data.
  • Scalability: While acknowledging ongoing experimentation, the paper notes that current triple store implementations can handle tens of billions of RDF triples in real-time, suggesting scalability for large datasets.

Limitations and Future Work

  • Technical Challenges: The authors acknowledge the need for further work on representing quantitative relationships (e.g., IC50 values) and provenance (data source and history) within RDF.
  • Ontology Development: Creating comprehensive, large-scale ontologies for complex domains like chemistry remains challenging. The paper suggests a hybrid approach of using existing ontologies and building domain-specific ones.
  • Data Quality: Ensuring the quality and comparability of data from diverse sources is a significant hurdle. The paper suggests addressing this at the tool level, allowing users to select datasets and understand their caveats.
  • Mapping to Real Problems: A key area for future work is mapping these integrative tools and algorithms effectively to specific drug discovery problems and demonstrating their value in accelerating the discovery of new drugs.
  • Emerging Data Sources: The paper anticipates that the massive increase in patient-level data (EMRs, genomics) will necessitate further advancements in triple-store and algorithm scalability.
  • Industry-Academia Collaboration: Continued collaboration is deemed essential for validating these methods and accelerating progress.

Summary updated

BibTeX
@inproceedings{wild2011systems,
  author = {David Wild and Ying Ding and Amit Sheth and Lee Harland and Eric Gifford and Michael S. Lajiness},
  title = {{Systems chemical biology and the Semantic Web: what they mean for the future of drug discovery research}},
  booktitle = {Drug},
  year = {2011},
  doi = {10.1016/j.drudis.2011.12.019},
}
Figure