Reduction in accuracy of genomic prediction for ordered categorical data compared to continuous observations

Background

Accuracy of genomic prediction depends on number of records in the trainingpopulation, heritability, effective population size, genetic architecture,and relatedness of training and validation populations. Many traits haveordered categories including reproductive performance and susceptibility orresistance to disease. Categorical scores are often recorded because theyare easier to obtain than continuous observations. Bayesian linearregression has been extended to the threshold model for genomic prediction.The objective of this study was to quantify reductions in accuracy forordinal categorical traits relative to continuous traits.

Methods

Efficiency of genomic prediction was evaluated for heritabilities of 0.10,0.25 or 0.50. Phenotypes were simulated for 2250 purebred animals using 50QTL selected from actual 50k SNP (single nucleotide polymorphism) genotypesgiving a proportion of causal to total loci of.0001. A Bayes Cπ threshold model simultaneously fitted all 50k markersexcept those that represented QTL. Estimated SNP effects were utilized topredict genomic breeding values in purebred (n = 239) or multibreed (n =924) validation populations. Correlations between true and predicted genomicmerit in validation populations were used to assess predictive ability.

Results

Accuracies of genomic estimated breeding values ranged from 0.12 to 0.66 forpurebred and from 0.04 to 0.53 for multibreed validation populations basedon Bayes C π linear model analysis of the simulated underlyingvariable. Accuracies for ordinal categorical scores analyzed by the Bayes Cπ threshold model were 20% to 50% lower and ranged from0.04 to 0.55 for purebred and from 0.01 to 0.44 for multibreed validationpopulations. Analysis of ordinal categorical scores using a linear modelresulted in further reductions in accuracy.

Conclusions

Threshold traits result in markedly lower accuracy than a linear model on theunderlying variable. To achieve an accuracy equal or greater than forcontinuous phenotypes with a training population of 1000 animals, a 2.25fold increase in training population size was required for categoricalscores fitted with the threshold model. The threshold model resulted inhigher accuracies than the linear model and its advantage was greatest whentraining populations were smallest.