Improving Genomic Prediction in Cassava Field Experiments Using Spatial Analysis

Area of Science:

• Agricultural Science
• Plant Breeding
• Genetics

Background:

• Cassava (Manihot esculenta) is a vital staple crop in sub-Saharan Africa.
• Accurate estimation of breeding values is crucial for developing improved cassava varieties.
• Field heterogeneity can significantly impact the accuracy of breeding trials.

Purpose of the Study:

• To investigate the impact of spatial variation on genomic selection models in cassava.
• To determine the optimal spatial kernel for improving prediction accuracy in cassava breeding.
• To enhance the predictability of breeding values by integrating spatial analysis into genomic selection.

Main Methods:

• Employed parametric spatial kernels (Power, Spherical, Gaussian) within a genomic selection framework.
• Fitted spatial and genomic kernels simultaneously to account for both spatial heterogeneity and genetic relatedness.
• Validated model predictability using 10-fold cross-validation, repeated five times, comparing models with and without spatial kernels.

Main Results:

• Incorporating spatial variation significantly increased the predictability of genomic selection models.
• Real data showed a median accuracy increase of 3.4%, while simulations demonstrated up to a 21% improvement.
• Gaussian spatial kernels, particularly in the row direction, were effective in explaining spatial variance in most scenarios.

Conclusions:

• Accounting for spatial heterogeneity is essential for increasing accuracy in cassava breeding value estimation.
• The integration of spatial kernels into genomic selection models offers a robust strategy for enhancing prediction accuracy.
• This approach is effective regardless of the heritability of the trait being studied.