Evaluation of Genomic Prediction Algorithms for Reducing Selection and Breeding Cycles in Shea Tree (Vitellaria Paradoxa)

The focus of this study was to determine the genomic prediction (GP) algorithms with the highest prediction accuracies for reducing the breeding and selection cycles in Vitellaria paradoxa. The efficiency of the GP algorithms were compared to evaluate five Shea tree growth traits in 708 genotypes with 30734 Single Nucleotide Polymorphic (SNPs) markers, which were reduced to 27063 after removing duplicates. Five hundred forty-nine (77.54%) Shea tree training population and 159 (22.46%) training population were genotyped for 30734 single nucleotide polymorphisms (SNPs) and phenotyped for five Shea tree growth traits. We built a model using phenotype and marker data from a training population by optimizing its genomic prediction accuracy for effectiveness of GS. The phenotype and marker data were used for cross validation of the prediction accuracies of the different models. Prediction accuracies varied among the genomic prediction algorithms based on the five phenotypic traits. We determined the best genomic algorithm that is more suitable for reduction of selection and breeding cycles in Vitellaria paradoxa. The GP algorithms were evaluated and we conclude that rrBLUP is the best for improving the prediction accuracy for reducing the breeding cycle in Shea tree.