Applying knowledge-driven mechanistic inference to toxicogenomics

Abstract

<jats:title>Abstract</jats:title><jats:p>Government regulators and others concerned about toxic chemicals in the environment hold that a mechanistic, causal explanation of toxicity is strongly preferred over a statistical or machine learning-based prediction by itself. Elucidating a mechanism of toxicity is, however, a costly and time-consuming process that requires the participation of specialists from a variety of fields, often relying on animal models. We present an innovative mechanistic inference framework (MechSpy), which can be used as a hypothesis generation aid to narrow the scope of mechanistic toxicology analysis. MechSpy generates hypotheses of the most likely mechanisms of toxicity, by combining a semantically-interconnected knowledge representation of human biology, toxicology and biochemistry with gene expression time series on human tissue. Using vector representations of biological entities, MechSpy seeks enrichment in a manually-curated list of high-level mechanisms of toxicity, represented as biochemically- and causally-linked ontology concepts. Besides predicting the canonical mechanism of toxicity for many well-studied compounds, we experimentally validated some of our predictions for other chemicals without an established mechanism of toxicity. This framework can be modified to include additional mechanisms of toxicity, and is generalizable to other types of mechanisms of human biology.</jats:p><jats:sec><jats:title>Author summary</jats:title><jats:p>Several recent computational methods have displayed excellent performance in predicting toxicity outcomes [1–3] of chemicals. Yet, to our knowledge, there is to date no computational approach to generate mechanistic hypotheses to answer <jats:bold>why</jats:bold> these chemicals elicit a toxic response. There is great value in understanding the mechanism of toxicity for a chemical that appears to elicit an adverse response. Novel small molecule development is one example, where a chemical that failed initial toxicological screenings could be assessed to evaluate the actual mechanism of toxicity, greatly reducing research time and expenses on subsequent ones. The value of a mechanistic awareness of toxicity also applies to pharmacovigilance, when researching rare adverse effects of a drug in subsets of the population. The development of oncological chemotherapeutics is another example, where certain mechanisms of cytotoxicity can actually be desirable to eliminate different types of tumor cells. More importantly, the costs, time expenditure, and ethical concerns of toxicity animal models, make in vitro and in silico approaches an enticing alternative. We present a solution that uses a combination of gene expression assays and biomedical knowledge to address the gap of answering the <jats:bold>why</jats:bold> question.</jats:p></jats:sec>

A Framework for Automated Construction of Heterogeneous Large-Scale Biomedical Knowledge Graphs

“…Our preliminary results evaluated different parameterizations of PheKnowLator KGs on a variety of dimensions. PheKnowLator KGs have also recently been used in applications of toxicogenomic mechanistic inference [14] and biomedical hypergraphs [19] and we'd like to invite the community to collaborate with us to examine the utility of PheKnowLator across a wider variety of use cases. Although not included in this preliminary evaluation, we are in the process of running different reasoners [20] over each KG parameterization.…”

“…To date, the vast majority of KG construction algorithms have been developed in order to create more manageable representations of large free-text corpora (e.g. scientific articles) [9,10] , to derive novel associations between existing concepts [11,12] , and add evidence to existing systems or KGs [13,14] . While many data-driven KG construction methods have been developed, they remain largely unable to automatically construct KGs from multiple disparate data sources, combine KGs created by different systems, and collaborate or share KGs across institutions due to their inability to account for the use of different schemas, standards, and vocabularies [15] .…”