| dc.description.abstract | Genomic selection (GS) in rubber tree (Hevea brasiliensis) has huge potential to meet future demands of
rubber in an economically and environmentally sustainable way. In Hevea breeding programmes,
genomic selection can be used early in the breeding pipeline to obtain genomic estimated genetic values
(GEGVs) for making clonal selections for further large-scale evaluation as potential commercial clonal
cultivars. Thus, genomic selection could enhance the efficiency of Hevea breeding significantly through
decreasing the generation interval and increasing selection intensity, therefore increasing genetic gains
per cycle. Within-family genomic selection for rubber latex yield was performed using two sets of 179
and 125 F1 clones from a cross between RRIM600 and PB260 evaluated in two separate phenotypic trials
in Côte d‘Ivoire. The clones were genotyped using the genotyping-by-sequencing (GBS) approach, which
resulted in 3,420 SNPs. A genetic linkage map of the rubber clones was constructed using the JoinMap
5.0 software and two marker imputation methods (Beagle 3.3 and random forest algorithm) were used to
impute the missing marker data. The ridge regression best linear unbiased prediction (rrBLUP) was used
to predict the GEGVs of clones across-sites. In addition, the effect of marker density on genomic
selection accuracy was investigated. Furthermore, the GS accuracies obtained were compared to the GS
accuracies obtained using SSR markers and the same phenotypic data. The genetic map contained 1,769
SNPs spanning 2600.9 Centimorgans (cM) and with an average of one SNP in every 1.47 cM. The
genetic map also encompassed 308 SSR markers which spanned across 18 linkage groups and with a
density of one marker in every 8.4 cM. Beagle imputation performed better than random forest imputation
(RFI) as it gave a GS accuracy of 0.52, against 0.48 with RFI. Results also showed that GS accuracy
increased with an increase in marker density, and a plateau was reached at 1,000 SNPs with Beagle
imputed marker data and at 2,000 SNPs with RFI marker data. The mean between site GS accuracy
obtained in this research is similar to the one obtained using SSR markers and the same phenotypic data,
opening the way to a cost-effective application of GS in rubber. Results of this study demonstrate that
GBS is a rapid, efficient and cost-effective approach for implementing genomics-assisted breeding. This
research also showed that GS has high potential to increase yield genetic gain in rubber breeding.
Key words: genomic selection, genomic estimated genetic values, genotyping-by-sequencing,
genetic gain, rubber tree. | en_ZW |
