Naslov
The prediction accuracy of genetic values is affected by imputation method within a wheat biparental population

Autori
Plavšin, Ivana ; Gunjača, Jerko ; Novoselović, Dario

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Book of Abstracts / - , 2021, 101-102

Skup
6th Conference on Cereal Biotechnology and Breeding

Mjesto i datum
Budimpešta, Mađarska, 03.11.2021. - 05.11.2021

Vrsta sudjelovanja
Poster

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
wheat quality ; genomic selection ; RR-BLUP ; marker density ; imputation

Sažetak
Genomic selection is one of the newly developed methods of marker-assisted selection in which molecular markers covering the whole genome are used. The main goal of genomic selection is to predict the genetic value of the individuals within validation population based on genomic estimated breeding value (GEBV). In our study, RR-BLUP model was used to predict grain quality traits within biparental wheat population. The population is comprised of 139 recombinant inbred lines (RILs) derived from a cross between Bezostaya-1 and Klara cultivars. The experiment was conducted during 3 consecutive years at Osijek and Slavonski Brod (Croatia). Phenotypic data for quality traits grain protein content (GPC), wet gluten content (WGC) and test weight (TW) were collected. Population was genotyped using the DArTseq technology and a total of 1087 SNPs were used for genomic selection. To assess the effect of marker density on prediction accuracy maximum number of markers (NM = 1087) and two subsets (NM = 544 and NM = 272) were used in predictions. Imputation of missing genotype data for each dataset was done using two approaches: mean imputation (MNI) and imputation within Beagle software, in order to assess the effect of the type of imputation of different marker densities on the prediction accuracy. Cross-validation was performed by randomly splitting the dataset into training and validation population in the ratio 80:20. Prediction accuracy was expressed as the mean value over the total number of cross- validation iterations which was set to 10 000. Among the quality traits examined, the highest prediction accuracy was obtained for TW, regardless of the NM size and imputation method used. As expected, reducing the NM had a large negative effect on prediction accuracy for all examined quality traits. The decrease in prediction accuracy was even more pronounced when a dataset imputed by Beagle software was used. The reduction of prediction accuracy with reducing NM was less severe when MNI was applied, suggesting that MNI may be a method of choice for imputation within a biparental population when the size of the available marker data is limited.

Izvorni jezik
Engleski

Znanstvena područja
Poljoprivreda (agronomija)

POVEZANOST RADA