Using visual scores and categorical data for genomic prediction of complex traits in breeding programs

Most genomic prediction methods are based on assumptions of normality due to their simplicity, robustness, and ease of implementation. However, in plant and animal breeding, target traits are often collected as categorical data, thus violating the normality assumption, which could affect the prediction of breeding values and the estimation of crucial genetic parameters. In this study, we examined the main challenges of categorical phenotypes in genomic prediction and genetic parameter estimation using mixed models, Bayesian approaches, and machine learning techniques. We evaluated these approaches using simulated and real breeding data sets. Our contribution in this study is a five-fold demonstration: (i) collecting data using an intermediate number of categories (1 to 3 and 1 to 5 scores) is the best strategy, even considering errors and subjectivity associated with visual scores; (ii) in the context of genomic prediction, Linear Mixed Models and Bayesian Linear Regression Models are robust to the normality violation, but marginal gains can be achieved when using Bayesian Ordinal Regression Models (BORM) and Random Forest Classification technique; (iii) genetic parameters are better estimated using BORM; (iv) our conclusions using simulated data are also applicable to real data in autotetraploid blueberry, which can guide breeders decisions; and (v) a comparison of continuous and categorical phenotype testing for complex traits with low heritability, found that investing in the evaluation of 600-1000 categorical data points with low error, when it is not feasible to collect continuous phenotypes, is a strategy for improving predictive abilities. Our findings suggest the best approaches for effectively using categorical traits to explore genetic information in breeding programs, and highlight the importance of investing in the training of evaluator teams and in high-quality phenotyping. Key messageAn approach for handling categorical data with potential errors and subjectivity in scores was evaluated in simulated and blueberry recurrent selection breeding schemes to assist breeders in their decision-making.