EMPIRICAL VALIDATION OF GENETIC GAINS OF GENOMIC SELECTION IN CASSAVA (Manihot esculenta Crantz) BREEDING
ABSTRACT
The success of a genetic breeding program can be assessed by the genetic gain obtained and successful varieties released. The validation of genomic selection (GS) efficacy in cassava (Manihot esculenta Crantz) breeding was the core objective of this study. A population of three cycles of genomic selection was evaluated to assess the genetic gains made in root yield, dry matter content and mosaic disease resistance in cassava. In the combined analysis, medium prediction accuracy was obtained for the dry matter content (r = 0.58), root weight (r = 0.49) and cassava mosaic disease resistance (r = 0.49) while lower accuracy was obtained for the root number. The expected gain and the realized selection gain for each of the traits are discussed here. Positive genetic gain (3.03%) was obtained in cycle two (C2) relative to and cycle one (C1), after an initial genetic loss (-1.62%) from cycle zero (C0) to C1 in the root weight. The translation of these genetic gains to actual root yield performance reflected in the top clones in the cycles of selection compared to the standard checks. The dry matter gain was highest (4.97%) in the C1 relative to C0; however, there was a decline in genetic gain in the subsequent cycles of selection. In a separate genomic prediction study, an assessment of prediction accuracy was conducted in a clonal evaluation trial, and this was compared with the accuracy of the prediction phase with genome-wide single nucleotide polymorphic markers, and the genomic estimated breeding values that were used for selecting the clones that were evaluated on the field. It was observed that prediction accuracy with genome-wide SNP markers has a direct relation to the field performance for all the root weight, dry matter content, harvest index and mosaic disease resistance. However, the cross validation from the field trial was lower than the cross-validation from the prediction with the molecular markers. It is imperative to continually find ways of improving the accuracy of genomic prediction models for traits of interest especially for a complex trait like root yield which has significant additive and non-additive gene actions in cassava if GS is to be fully implemented in cassava. A high prediction accuracy was obtained for the plant types and architecture traits which gives the flexibility of selecting good plant types even before the actual phenotyping. These findings suggest that genomic selection is a feasible complementary breeding scheme to conventional breeding especially in population improvement in cassava, hence the integration of GS with the existing conventional breeding programmes would greatly enhance genetic improvement of traits with greater efficiency and timely development of cassava varieties for the end-users.