Random Forest-based genomic prediction allows selection of barley genotypes from the variety collection for inclusion evaluation in distinctness, uniformity, and stability (DUS) trials
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Shortening the time taken to release a new crop variety could improve the responsiveness of the plant breeding sector. To gain Plant Breeders’ Rights, and for certain marketing requirements, candidate varieties must be distinct from all known varieties in common knowledge (the variety collection) through assessment against Distinctness, Uniformity and Stability (DUS) criteria. The assessment uses field comparisons of morphological DUS characteristics between the candidate and its most similar varieties from the variety collection. Similar varieties cannot be identified immediately as candidate varieties are received without characteristic data. Here, we investigated if genomic prediction could facilitate most similar variety selection. A barley training population was assembled with 1,171 genotypes, covering ~70% of the UK barley variety collection at the time of study. Genotyping was undertaken to obtain 36,736 genetic markers and phenotypes were compiled for 28 DUS characteristics. Genomic prediction was completed using Random Forest (RF) for classification and Ridge Regression Best Linear Unbiased Prediction (rr-BLUP). RF typically showed the highest proportion of correctly classified calls and marginally higher prediction accuracies for characteristics with high heritability and fewer characteristic notes. Prediction accuracies varied more between methods for characteristics with lower heritability and more notes, where rr-BLUP performed best. Weighting mapped genetic loci in the RF prediction improved prediction accuracies. Comparisons made with historic data showed strong overlap between the most similar varieties identified via predicted or observed characteristics, indicating distinctness comparisons could begin a year earlier. Therefore, the combination of biomolecular technology and genomic prediction could enable faster distinctness assessment.
