As toxicology begins to move towards high-throughput and high-content screening, scientists are becoming deluged with data on the effects of chemicals at a molecular/mode-of-action level. This information has the potential to be used for prediction of adverse effects at an organismal level, but these predictions would be facilitated by systematic organisation of data by presumed mode of action (MOA). Such a systematic organisation (an ontology) would (1) provide linkage of molecular data to traditional toxicology study outputs and to human disease states, (2) provide clarity as to whether existing high-throughput or high-content approaches are sufficiently inclusive of the universe of MOAs for toxicity, and (3) serve as an organising structure for constructing adverse outcome pathways (AOPs). Ontologies provide one means to deal with such data in a structured manner, while also providing a mechanism for integrating with larger IT-infrastructures to facilitate decision making.
Ontologies are often used when data from disparate sources need to be integrated. This allows investigators to make complicated queries of the data encoded by the ontology. Ontologies can be foundational, forming the basis upon which other ontologies will be built, or application-specific, where there is a specific scope or purpose for the ontology. Although ontologies use a controlled vocabulary (i.e. a distinct set of words is used to describe each concept), it is the relationship between concepts that sets it apart from a controlled vocabulary and make it useful for computing and learning.
Scientists will appreciate the ability of ontologies to quickly build hypotheses that can be followed up. The ontology may even be able to provide insight that can be used in a decision-making system for evidence integration. For example, the ontology may hypothesise that a particular chemical causes issues with eye development; this leads to experimental follow-up and those results can also be managed in the ontology. All of this, and other information, could be placed in a decision-support tool, which integrates additional information, such as historical information that this class of chemicals causes developmental toxicity at high doses that are not seen in other vertebrates. All of the computational aspects can be fully automated, allowing decisions to be made more quickly, while managing a larger portfolio of chemicals. A similar set-up combining the ontology with a decision-support tool can be used for other applications, such as streamlining the evidence integration process in various risk assessments.
Ontologies also allow for the automated prediction of AOPs. Once the ontology contains several AOPs, there may be interconnections, where common key events (KEs) exist. This allows for the automated construction of AOP networks. As the number of AOPs grows, the complexity of the network will also grow, generating novel AOPs that have not been previously explored. These computationally predicted AOPs (cpAOPs) are new AOP hypotheses that can be followed up experimentally. The experimental results can be used to fine-tune/prune relationships within the network, potentially graduating an AOP from being a cpAOP to a putative or accepted AOP.
This report focuses on the need to explore the field of developmental toxicology in this way, by creating a formal system (i.e. ontology) that organises the knowledge of chemical structure, developmental biology and developmental toxicology so that it becomes possible to predict and explain which chemicals are likely to induce human developmental toxicity. The authors believe this is needed for the following reasons:
- It would overcome some of the limitations of current safety testing by exploiting the state of the science and the increasing amounts of data that can inform us about MOAs that lead to adverse outcomes.
- It would improve public health protection through increased relevance and accuracy of testing.
- It would facilitate the design of pharmaceuticals and other chemicals so that they are unlikely to have the potential for developmental toxicity in humans.
It would save resources (time, animals).