Special Report 19


Here we can move descriptions into related defects based on the embryology of the target system. In this way, the unique view of DevTox as an observation-based ontology system is extended with new concepts and relations derived from an embryology-based ontology. What is most useful is to map apical endpoints to developmental ontology that gives order and timing to pathogenesis. Descriptions are first integrated into elementary concepts (one and only one ‘parent’ and distinctive ‘children’: “necessary” must occur in order to define the relationship, and “sufficient” may be enough to define the relationship). For example, a ‘disproportionate reduction in size of the optic globe’ is an elementary concept; it is a necessary criterion for microphthalmia, but not sufficient because complete apparent anophthalmia may also apply. More information is needed to decide between these phenotypes (e.g. some evidence of optic tissue).

The medical ontology for human birth defects does not have the necessary specificity in this regard. Consider the classification system for human malformations adopted for the National Birth Defects Prevention Network, based on The Metropolitan Atlanta Congenital Defects Program (MACDP) (Correa-Villaseñor et al., 2003). This is a population-based, birth defects surveillance program motivated by the thalidomide tragedy. It tracks approximately 50,000 births per year for 35 years to monitor trends over time and find co-occurrence patterns that elucidate etiology. Information collected on each infant for over 100 individual defects is classified using a six-digit code modified from the British Paediatric Association (CDC-BPA codes) based on the WHO International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM codes). Codification queries the databases by anatomic specificity and medical classification of defects and is driven by the need to assess prevalence rates (e.g. cases per 10,000 births) and syndromes (e.g. CHARGE). Although CDC-BPA and ICD-9-CM meet the needs of large, population-based birth defects surveillance programs, the classification is not linked to a developmental ontology. As such, these classifications do not systematically address embryology.

Text-based systems such as the Medical Dictionary for Regulatory Activities (MedDRA) have been used to group human birth defects by anatomical location or clinical condition in smaller databases. A related classification system proposed for signalling teratogenic clusters has basically condensed and rearranged the familiar CDC-BPA codes into a three-tiered hierarchy: <organ system>, <preferred defect term>, and <reported defect term> (Scheuerle and Tilson, 2002). The first level uses traditional medical reporting, modified where necessary to impose current embryological considerations to 20 basic categories of organ systems and specified syndromes. The second level attempts to define defects by the most standard and recognisable terminology as defined in the diagnosis, which is the third level. By aggregating individual defects into pathogenic groups, the system improves diagnostic specificity for multiple terms used to describe the same defect. A three-tiered system allows the database to be queried at different levels of specificity to make inferences across smaller population sizes. Although this can increase the visibility of an early signal of developmental toxicity, it is still basically an anatomy-based observational system.

An example of what can be done is the work of Georgas et al. (2015) who have developed a definitive spatiotemporal description, at the level of organ, tissue and cell type, for the developing lower urinary and reproductive tracts in the mouse. The information has been incorporated into a text-based anatomical ontology spanning developmental time, space and gender.

Formalising associative relationships between anatomical structure and spatial location, functional system and chronological stage in the embryo requires hierarchical information. Bard (2005) discussed the difficulties imposed in trying to formalise abnormal anatomy across organisms.