Comparison of Statistical Methods and Reliability of Genomic Prediction within and Between Environments and Breeding Populations in Coffea canephora

Simulation and empirical results have been shown that genomic predictions present sufficient accuracy to guarantee success in breeding programs. Although some crops have been benefited from this contemporary approach, studies in the Coffea genus are still in their infancy. Until the present, there is no evidence supporting how predictive models works across populations and environments or, even, their performance for different complex traits. Considering that predictive models are based on biological and statistical assumptions, it is expected that their performance varies depending on the true underlying genetic architecture of the phenotype. Real data from two experimental populations of Coffea canephora, evaluated in two environments (sites) and SNPs identified by Genotyping-by-Sequencing (GBS) were used to investigate the genotype-phenotype relationship. Nine Bayesian models, with different prior distributions for the marker effects, and three regularized linear regression models were considered. The predictive abilities were assessed using a Replicated Training-Testing evaluation, with 30 repetitions. Correlation between predicted and observed values, mean squared prediction error and computational time were used to compare the model performances. In addition, SNP effects were investigated, in order to learn about underlying biology related to genomic regions affecting the phenotype and their interactions. For the four traits evaluated, minimal differences were observed across the models in terms of predictive accuracy. A slight advantage of Bayesian methods was observed, although more computationally-demanding was required. Predictions within-population, in average, were more accurate than between populations. Biological insights revealed genetic variant with specific signals within populations and environments, as well as, variant with generalist signals. As consequence, these results have potential to reshape traditional breeding programs, including genomic predictions on the breeding schemes. Further, it opens important perspectives to investigate these SNPs in more details.