Cassava yield traits predicted by genomic selection methods.

Area of Science:

• Plant breeding and genetics
• Quantitative genetics
• Genomics

Background:

• Genomic selection (GS) offers an alternative to traditional phenotypic selection for crop improvement, especially when traits are difficult or costly to measure.
• Cassava (Manihot esculenta Crantz) breeding programs can benefit from GS to accelerate genetic gains.
• Understanding the influence of population structure is crucial for the reliable application of GS.

Purpose of the Study:

• To evaluate the predictive ability and consistency of different GS methods for cassava yield and dry matter content.
• To assess the impact of population structure on GS accuracy in cassava.
• To compare the performance of BayesB, BLASSO, RR-BLUP, G-BLUP, and RKHS methods.

Main Methods:

• Obtained Best Linear Unbiased Predictors (BLUPs) and deregressed BLUPs for 888 cassava accessions.
• Utilized a 48K single nucleotide polymorphism dataset for genomic predictions.
• Applied five GS methods (BayesB, BLASSO, RR-BLUP, G-BLUP, RKHS) and four cross-validation strategies accounting for population structure.

Main Results:

• Dry matter content prediction was more accurate (0.5655-0.5670) than fresh (0.4569-0.4756) and dry root yield (0.4689-0.4818).
• The Reproducing Kernel Hilbert Spaces (RKHS) method showed higher efficiency and consistency for yield traits across validation scenarios.
• High correlations (0.99-1.00) among GS methods indicated similar clone selection outcomes.

Conclusions:

• GS is effective for cassava breeding, with higher accuracy for dry matter content than yield traits.
• RKHS demonstrated robust predictive performance for yield traits.
• Population structure significantly affects GS predictions, emphasizing the importance of relatedness between training and validation populations for successful implementation.