Knowledge representation for curation and analysis of crop phenotypic data underpinning nutritional security

Liliana Andrés Hernández

doi:10.25918/thesis.157

Achieving global and regional nutritional security presents many challenges, including maximizing dietary inputs from crop plants. Comparison of nutritional composition between crops, cultivars and associated plant genetic resources requires access to datasets that provide insights into genetic and environmental sources of variation. At present, the ability to access and compare crop phenotypic datasets is limited, particularly for traits relating to dietary nutrition. To date, the use of common trait descriptors or formal systems of knowledge representation such as controlled vocabularies or ontologies has been fragmentary. The primary aim of this thesis is to describe the evaluation and development of ontologies that represent information about crop phenotypic traits in compliance with the Findable, Accessible, Interoperable, and Reusable (FAIR) principles. This provides the opportunity to compare approaches in generating a generalised trait ontology applicable to many crops. Generic efforts to standardize the description and management of crop trait data are critically reviewed, and several issues identified, such as inconsistencies in the syntactic and semantic crop trait descriptions that currently constrain their exchange and comparison. Bambara groundnut (Vigna subterranea) was used as an exemplar underutilised minor crop to assess available formal frameworks to curate trait datasets so that they are accessible for comparative analysis. The challenges in assembling and using a crop-specific Trait Dictionary (TD) are outlined, and the extent to which any Crop Ontology (CO) derives knowledge from existing ontological definitions and relationships quantified. Evaluation of syntactic and semantic cohesiveness of trait descriptors within the Crop Ontology (CO) system helped to identify generic issues limiting reuse of trait names for comparative analysis of major and minor crops. These included inconsistencies in trait names assigned within the 28 published CO:TDs, a relative lack of cross-referencing to other ontologies, and a flat ontological structure for classifying traits. The Trait Dictionary (CO_366) was generated for bambara groundnut, and formal terms were associated with trait descriptors and experimental datasets. These were curated to be compliant with the Minimal Information About a Plant Phenotyping Experiment (MIAPPE) metadata standard, and made available within the publicly accessible CropStoreDB database. For the representation of dietary nutritional information, the Crop Dietary Nutrition Ontology (CDNO) was first proposed and then developed to provide a more comprehensive vocabulary compatible with terms and concepts widely used within national and international Food Composition Databases (FCDBs). The CDNO represents a hierarchical structured vocabulary with two major classes ‘dietary nutritional components’ and ‘nutritional component concentration’, developed by reusing terms from the Chemical Entities for Biological Interest (CHEBI), the Environmental Ontology (ENVO), Phenotype and Trait Ontology (PATO), Relation Ontology (RO) and the Plant Ontology (PO). The reuse of terms from existing ontologies created terms categorised as pre-composed (pre-coordinated), which follows the Open Biological and Biomedical Ontology (OBO) principles for establishment of logically well-formed, scientifically accurate and interoperable controlled vocabularies. The CDNO is open and will be releasing versions constantly with new updates being track in the GitHub repository. The CDNO involved collaboration and co-development of terms with the Food Ontology (FoodOn) for the representation of organismal materials terms, so that they could be assigned in different combinations as post-composed terms with the CDNO ‘nutritional component concentration’ terms. To demonstrate how the CDNO can assist in management and navigation of crop datasets that quantify concentration of chemical components, specific use case implementations are presented. The use case describes the association of existing trait descriptors for bambara groundnut with CDNO terms incorporated into an axiom that reuses terms from FOODON, PO and NCBI organismal classification ontology (NCBITaxon). A more systematic and inclusive approach was proposed for a second-generation crop phenotypic trait ontology system that would meet the challenges of data integration and downstream analysis, by ensuring consistency in knowledge representation and implementation of machine-readable axiomatic definitions. This would require establishing a universal, systematic and deeper set of ontology sub-classes reusing terms from the Plant Trait Ontology (TO) and other relevant ontologies. In conclusion, generic systems of knowledge representation for crop plant phenotypes are expected to facilitate curation of datasets according to FAIR principles. Systems such as CDNO should help increase informed dialogue between crop scientists, breeders and other stakeholder in agricultural production and food supply, including dieticians and nutritionists.

Knowledge representation for curation and analysis of crop phenotypic data underpinning nutritional security

Files and links (1)

Metrics

Abstract

Details

Knowledge representation for curation and analysis of crop phenotypic data underpinning nutritional security

Files and links (1)

Metrics

Abstract

Details

Southern Cross University Social media