Breeding Jatropha By Genomic Selection: A Pilot Assessment of Accuracy of Predictive Models

Jatropha is an oilseed crop with great potential for biofuel production. Given its long breeding cycle (BC) and late expressing traits, Genomic Selection (GS), could be a powerful approach to increase breeding efficiency. To evaluate predictive accuracy and potential impact of GS on Jatropha breeding, a pilot set of individuals (79) were phenotyped for the weight of 100 grains and genotyped with the DArTSeq platform (747 SNPs and 2.724 presence/absence variants - PAV). Predictive models were built using RR-BLUP , using a set of 47 (60%) randomly selected individuals as the training population (TP), while the remaining individuals were used as validation population (VP). Accuracy of phenotypic selection (PS), using the traditional REML/BLUP approach, was used as benchmark. Trait heritability was 0.37 while the accuracy of phenotypic selection was estimated at 0.60. GS accuracies following cross validation were 0.63 and 0.71 when using SNP or PAV markers, respectively. GS accuracies matched the benchmark PS accuracy, despite the limited training set, possibly due to a considerable effect of family relationships. If this holds true for other important traits, genomic breeding values may be estimated at the seedling stage, reducing Jatropha BC in at least five years (7 yr BC with GS vs 12 yr BC without GS). Being selection response inversely proportional to BC length, we calculated the expected impact of GS on Jatropha breeding. Considering the accuracies and cycles here reported, GS with may increase selection efficiency in Jatropha breeding by over 100%.