The impact of epistasis on the genomic prediction of non-assessed and non-overlapping single crosses in multi-environment trials
Year: 2025
Type: preprint
Abstract: The objective of this simulation-based study was to assess the impact of epistasis on the genomic prediction of untested single crosses (SCs) in multi-environment trials (METs), assuming seven types of digenic epistasis. We genotyped two groups of 70 doubled-haploid lines and phenotyped all 4900 SCs in five environments. The average density for SNPs was 0.03 cM. Regarding the distribution of the SCs over environments, we adopted 10 and 30% of tested SCs and 80% of non-overlapping SCs. To assess the efficacy of the genomic prediction we computed coincidence index (CI) and prediction accuracy. The percentage of the genotypic variance due to epistasis ranged from 18 to 48%. The CI ranged from 0.24 to 0.38, under the lower training set, and between 0.30 to 0.50, assuming the higher training set. Fixing complementary epistasis and increasing the ratio epistatic variance/genotypic variance from 18 to 39% led to a decrease in the CI in the range 6.5 to 22.0%. Accuracy showed a highly positive correlation with CI. Assuming epistasis but fitting the additive-dominance model led to a decrease in the CI. Epistasis can significantly affect the genomic prediction of SCs in METs, depending on type and proportion of the genotypic variance due to epistasis. The prediction accuracy efficiently expresses the efficacy of genomic prediction of untested SCs. The genomic prediction of untested SCs in each environment and across environments is very effective, even under a low training set size.
Authors: José Marcelo Soriano Viana, Jean Silva
Institution: Universidade Federal de Viçosa
Language: English
Subfield: Genetics
Domain: Life Sciences
SDG: Zero hunger
Open Access status: gold
OpenAlex ID: W4413279859