Ontology

Full training materials about DeltaBreed Servers can be found in text form can be found at this link: Ontology.

Highlights from that training as well as videos can be accessed on this webpage.

Learning Targets for this training module:

  • Why your program needs an ontology.
  • How to get started building an ontology.
  • How to use the ontology template to upload a list of ontology terms.
  • How to add an individual ontology term.
  • How to search for ontology terms.

Trait Ontology is a controlled vocabulary used to describe and codify a collection of phenotypic traits in breeding programs. In DeltaBreed, ontology is defined as a controlled vocabulary to support the standardization of observation variables (or ontology terms) so that breeding data are unambiguous, interoperable and reusable. DeltaBreed follows the cropontology.org conceptual model that defines variables as a combination of a trait, a method and a scale. Note: breeders may use the term “trait” interchangeably with observation variable or phenotype, but DeltaBreed uses the term “trait” more specifically and the ontology is designed to manage any type of observation variable, not just phenotypes.

Setting and using Trait Ontologies are vitally important to allow:
• Databases to correctly store and use phenotypic data across years/trials/locations.
• Consistent collection of phenotypic data from person-to-person and program-to-program.
• Sharing and interpretation of data and resources between breeders and breeding programs.

How do I get started building an ontology for my program?

Reach out to your BI Coordinator! They are a great resource to help you get started. Also, consider working with other breeding programs of the same species to see if they already have a working ontology.
Use the following resources to see if they might help you define some ontology terms your breeding program often uses.

 

About Categorical Ontology Variables

When phenotyping, quantitative variables should always be a breeder preference, because quantitative measures are the best way to characterize quantitative traits and support parametric statistical inference. DeltaBreed supports both qualitative (nominal) and quantitative (ordinal) categorical variables. Nominal variables are described by 1-n categories. Ordinal variables are described by 2-n categories with their corresponding value ranges in DeltaBreed.

Best Practice:

  • When phenotyping use a quantitative variable when possible.
  • If a categorical phenotype is quantitative, create an ordinal variable.
  • To avoid ambiguity you must define both the category and the value range ordinal traits.
  • If you having a hard time assigning values to your categories, consider using fewer categories. Using fewer categories (<6) has been shown to be a good compromise between accuracy and error.
  • If you are curating an historical dataset, where values are unknown for an ordinal categories, you will need to create a nominal variable. Once you have loaded the historical data, consider archiving this variable and creating a new ordinal variable to use moving forward.

Important Considerations for Ontology Best Practices for Ordinal Scales.
The DeltaBreed ontology template requires a scale class for an ordinal ontology trait. There is debate amongst breeders and scientist as to whether the 1-9 ordinal scale is the most accurate way to define this type of unit. Although DeltaBreed allows for a 1-9 scale, we suggest best practice might be something different.
The severity of disease obtained with a qualitative ordinal rating scale is a rank-ordered numeric variable, but the qualitative ordinal rating scale is based on descriptions of symptoms (Madden et al. 2007; Agresti 2010). It is not statistically appropriate to take means based on these scales (Stevens 1946) as it has little meaning biologically, and violates assumptions underlying parametric tests. Qualitative ordinal scales can be analyzed using non-parametric statistics suitable for various experiment designs and distribution functions (Shah and Madden 2004; Fu et al. 2012).
This quote comes from “Understanding the ramifications of quantitative ordinal scales on accuracy of estimates of disease severity and data analysis in plant pathology.” Kuo-Szu Chiang and Clive H. Bock, published July 13, 2021

 

“The higher rates of misclassifications are probably affecting the use of 1–9 scores, making this category less efficient. Overall, we observed that using 1–3 and 1–5 scales for categorizing continuous data showed a good compromise between accuracy and error.
This quote comes from “Using visual scores for genomic prediction of complex traits in breeding programs.” Camila Ferreira Azevedo, Luis Felipe Ventorim Ferrão, Juliana Benevenuto, Marcos Deon Vilela de Resende, Moyses Nascimento, Ana Carolina Campana Nascimento, Patricio R. Munoz, published December 15, 2023

Ontology Template

Pay special attention to the “Readme” and “Example” tab within the Ontology Template. These tabs give a user important information that they will need in order to successfully upload a batch file or individual ontology terms.

Ontology Template

Uploading a batch file of ontology terms is the easiest way for a user to get all of the observational terms into DeltaBreed.

Upload ontology
Add individual terms, share ontology with other same species breeding programs and download an ontology

Users can also upload individual terms or download entire ontologies using DeltaBreed’s user friendly UI. Same species breeding programs can also choose to share ontologies.

Back to the main Learning DeltaBreed page
Next training module: Experiments, Observations, and Data Collection