Optimizing whole-genomic prediction for autotetraploid blueberry breeding

Abstract

Blueberry (Vaccinium spp.) is an important autopolyploid crop with significant benefits for human health. Apart from its genetic complexity, the feasibility of genomic prediction has been proven for blueberry, enabling a reduction in the breeding cycle time and increasing genetic gain. However, as for other polyploid crops, sequencing costs still hinder the implementation of genome-based breeding methods for blueberry. This motivated us to evaluate the effect of training population sizes and composition, as well as the impact of marker density and sequencing depth on phenotype prediction for the species. For this, data from a large real breeding population of 1804 individuals were used. Genotypic data from 86,930 markers and three traits with different genetic architecture (fruit firmness, fruit weight, and total yield) were evaluated. Herein, we suggested that marker density, sequencing depth, and training population size can be substantially reduced with no significant impact on model accuracy. Our results can help guide decisions toward resource allocation (e.g., genotyping and phenotyping) in order to maximize prediction accuracy. These findings have the potential to allow for a faster and more accurate release of varieties with a substantial reduction of resources for the application of genomic prediction in blueberry. We anticipate that the benefits and pipeline described in our study can be applied to optimize genomic prediction for other diploid and polyploid species.

Fig. 1. Predictive abilities and standardized mean squared error (MSE) values estimated for fruit firmness, fruit weight, and yield.

Results obtained considering two scenarios: a under cumulative increase of markers and b under cumulative increase of the number of probes. Letters on top of boxplots represent the results obtained in the post hoc analysis considering Tukey correction and σ = 0.05, groups that share a letter are not significantly different from one another and $a>b>\dots >z$.

Fig. 2. Percentages of the variance explained for the first component of the principal component analysis (PC) performed for the relationship matrices built for each marker density scenario.

Letters on top of boxplots represent the results obtained in the post hoc analysis considering Tukey correction and σ = 0.05, groups that share a letter are not significantly different from one another and $a>b>\dots >z$.

Fig. 3. Predictive ability obtained for fruit firmness, fruit weight, and yield when considering training population size and composition.

a Cumulative increase of the training population size considering family information and b cumulative increase of the training population size considering random sampling. Letters on top of boxplots represent the results obtained in the post hoc analysis considering Tukey correction and σ = 0.05, groups that share a letter are not significantly different from one another and $a>b>\dots >z$. The post hoc test was performed for each trait comparing all results for both sampling scenarios.

Fig. 4. Predictive ability values obtained in genomic prediction analyses for fruit firmness, fruit weight, and yield when considering six sequencing depth scenarios.

Letters on top of boxplots represent the results obtained in the post hoc analysis considering Tukey correction and σ = 0.05, groups that share a letter are not significantly different from one another and $a>b>\dots >z$.

Fig. 5. Predictive ability values obtained for fruit firmness, fruit weight, and yield when considering training population size, and sequencing depth.

a cumulative increase of the training population size considering random sampling; and b cumulative increase of the training population size considering family information. Letters on top of boxplots represent the results obtained in the post hoc analysis considering Tukey correction and σ=0.05, groups that share a letter are not significantly different from one another and $a>b>\dots >z$.

Fig. 6. The effect of probe and sequencing depth on prediction.

Predictive ability obtained for fruit firmness, fruit weight, and yield when considering probe density and average sequencing depth.