MechSpy

MechSpy infers mechanistic hypotheses of chemical toxicity by integrating gene expression time series data with a semantically-interconnected biomedical knowledge graph.

Key Features:

• Mechanistic Hypothesis Generation: Generates mechanistic hypotheses of toxicity by integrating high-level biological, toxicological, and biochemical knowledge with empirical gene expression data from human tissues.
• Knowledge Integration: Employs a semantically-interconnected knowledge representation of human biology, toxicology, and biochemistry and a manually-curated list of ontology concepts that are biochemically and causally linked.
• Vector Representations: Utilizes vector representations of biological entities to facilitate the search for enriched mechanisms within the knowledge base.
• Enrichment Analysis: Performs enrichment analysis over ontology concepts to identify pathways and mechanisms likely responsible for observed toxic responses.
• Experimental Validation: Has experimental validation for several chemicals, demonstrating prediction of canonical mechanisms for well-studied compounds and generation of novel hypotheses for others.
• Customizability and Generalization: Allows incorporation of additional mechanisms of toxicity and is generalizable beyond toxicology to other aspects of human biology research.

Scientific Applications:

• Toxicological Research: Provides mechanistic insights into chemical toxicity and helps prioritize mechanisms for further investigation, potentially reducing reliance on animal models.
• Pharmacovigilance: Supports analysis of rare adverse drug effects in specific population subsets by elucidating underlying toxicological mechanisms.
• Drug Development: Assesses mechanisms of toxicity for novel small molecules that fail initial screenings to inform downstream research decisions.
• Oncology: Predicts cytotoxic mechanisms of oncological chemotherapeutics to inform development of cancer treatments.

Methodology:

Integrates gene expression time series with a semantically-interconnected knowledge graph, represents biological entities as vectors, and performs enrichment analysis of manually-curated ontology concepts to identify enriched pathways and mechanisms.