Prospects for Genomic Selection in Cassava Breeding.

Area of Science:

• Agricultural Science
• Genetics
• Plant Breeding

Background:

• Cassava is a vital tropical staple crop, crucial for food security.
• Genomic selection (GS) is being adopted in African cassava breeding programs to accelerate development cycles.
• Previous studies indicated promising predictive abilities for GS in cassava.

Purpose of the Study:

• To evaluate the accuracy of seven prediction models across three scenarios: within-population, cross-population, and cross-generation prediction.
• To assess the impact of training population (TP) size on prediction accuracy, using random and genetic algorithm-based selection.
• To provide insights for optimizing GS strategies in cassava breeding.

Main Methods:

• Assessed prediction accuracy for seven traits using seven models in three scenarios.
• Investigated the effect of increasing training population size through random and genetic algorithm-selected phenotyping.
• Compared within-population, cross-population, and cross-generation prediction accuracies.

Main Results:

• Non-additive models showed a 10% improvement in prediction accuracy during cross-validation.
• Cross-population prediction accuracy was generally low (mean = 0.18), but improved for cassava mosaic disease when populations were combined.
• Increasing TP size was more impactful than model selection for improving accuracy across generations.
• Genetic algorithm selection achieved equivalent accuracy with one-third of the phenotyped progeny compared to full phenotyping.

Conclusions:

• Genomic selection shows potential for accelerating cassava breeding, particularly for traits like dry matter content and mosaic disease severity.
• Effective GS requires growing training populations, but phenotyping can be optimized by selecting subsets of individuals, reducing the phenotyping burden.
• Genetic algorithm-based selection offers a more efficient approach to phenotyping for increasing training population size.