Norman Munyengwa, Vincent Le Guen, Hermine Ngalle Bille, Livia M Souza, André Clément-Demange, Pierre Mournet, Aurélien Masson, Mouman Soumahoro, Daouda Kouassi, David Cros
Genomics 2021 MarGenotyping-by-sequencing (GBS) provides the marker density required for genomic predictions (GP). However, GBS gives a high proportion of missing SNP data which, for species without a chromosome-level genome assembly, must be imputed without knowing the SNP physical positions. Here, we compared GP accuracy with seven map-independent and two map-dependent imputation approaches, and when using all SNPs against the subset of genetically mapped SNPs. We used two rubber tree (Hevea brasiliensis) datasets with three traits. The results showed that the best imputation approaches were LinkImputeR, Beagle and FImpute. Using the genetically mapped SNPs increased GP accuracy by 4.3%. Using LinkImputeR on all the markers allowed avoiding genetic mapping, with a slight decrease in GP accuracy. LinkImputeR gave the highest level of correctly imputed genotypes and its performances were further improved by its ability to define a subset of SNPs imputed optimally. These results will contribute to the efficient implementation of genomic selection with GBS. For Hevea, GBS is promising for rubber yield improvement, with GP accuracies reaching 0.52. Copyright © 2021 Elsevier Inc. All rights reserved.
Norman Munyengwa, Vincent Le Guen, Hermine Ngalle Bille, Livia M Souza, André Clément-Demange, Pierre Mournet, Aurélien Masson, Mouman Soumahoro, Daouda Kouassi, David Cros. Optimizing imputation of marker data from genotyping-by-sequencing (GBS) for genomic selection in non-model species: Rubber tree (Hevea brasiliensis) as a case study. Genomics. 2021 Mar;113(2):655-668
PMID: 33508443
View Full Text