A genomic estimated breeding value in cannabis, often shortened to GEBV, is a model-based prediction of the inherited value a plant may pass to offspring for a defined trait. It combines genome-wide marker data with a training population that has reliable phenotype records. A GEBV can help rank untested or early-stage candidates, but it isn’t a measurement of overall quality and it doesn’t guarantee a descendant’s performance.
What the Estimate Represents
The word breeding matters. The estimate concerns transmissible genetic merit, not everything visible in the individual plant. Environment, management, developmental stage, measurement error, and non-additive effects can all contribute to phenotype without being passed to offspring in the same way.
The site’s explanation of heritability provides the essential context: genetic signal depends on the population and environments used to estimate it.
How a GEBV Is Built
- Define the trait, units, stage, and target production environment.
- Collect consistent phenotype records from a representative training population.
- Genotype that population across the genome and apply documented quality filters.
- Fit a model that connects marker relationships with the observed trait.
- Validate predictions on plants or families excluded from model fitting.
- Use the locked model to predict new candidates from their genotype data.
For a quantitative trait, the model may capture many small contributions rather than a single decisive locus. That’s why the training population and phenotype quality often matter more than a visually impressive marker panel.
GEBV, Phenotype, and EBV
A phenotype is an observed value on the plant. A traditional estimated breeding value may use phenotype and pedigree relationships. A GEBV uses genome-wide relationships or marker effects, often alongside phenotype, pedigree, environment, or repeated records. The label shouldn’t be used unless the method and inputs are clear.
Accuracy Is Part of the Result
A responsible report keeps prediction accuracy or reliability beside the GEBV. Accuracy depends on training size, trait heritability, marker density, relatedness between training and prediction candidates, population structure, and environmental representation. It can fall when the model moves into unrelated germplasm or a production system it hasn’t seen.
Genotype-by-environment interaction is especially important when a trait changes rank across sites. A single GEBV may be inadequate if the breeding objective needs environment-specific performance.
A Practical Reading Card
When comparing genomic breeding values, ask:
- Which exact trait and units were modeled?
- Which population and environments trained the model?
- How closely are these candidates related to that population?
- Was validation performed on withheld plants, families, or later cycles?
- What accuracy or reliability accompanies each estimate?
- Does the model version match the current report?
How to Use the Ranking
Use a GEBV as one input to parent selection, then review identity, fertility, disease status, diversity, complementary traits, and mating design. Avoid selecting several close relatives simply because they share the highest predictions. A breeding program needs genetic gain and enough diversity to keep making progress.
The Bottom Line
A GEBV is useful when it makes an expensive or slow decision earlier without pretending uncertainty disappeared. Keep the trait definition, training population, validation result, model version, accuracy, and candidate relatedness attached to every score. If those details are missing, you have a ranking but not enough evidence to know where it applies.
