Predicting Genetic Variance from Genomewide Marker Effects in Maize

Abstract

Predicting the genetic variance (VG) in a biparental population has been difficult. An effective procedure for predicting VG would likely require modeling of progeny segregation within each cross. Our objective was to determine whether the population mean, VG, and mean of the top 10% of progeny in a cross can be predicted effectively from genomewide marker effects. Eight maize (Zea mays L.) crosses that differed in predicted mean and VG were evaluated for plant and ear height, and growing degree days to silking across three locations in Minnesota in 2017. Each cross was represented by 120 to 144 random F3 lines. Correlations between the observed and predicted means of each breeding population were significant (P = 0.05) for all three traits (0.91 for plant height, 0.83 for ear height, and 0.80 for silking date). However, correlations between the observed and predicted VG were nonsignificant, ranging from -0.24 to 0.14 for the three traits. Correlations between the observed and predicted mean of the top 10% of progeny in each cross were significant for plant height (0.72) but not for ear height (0.56) and silking date (0.37). These results for predicting the mean of the top 10% of progeny reflected the ability to predict the mean but not VG. We concluded that while the means of breeding populations can be predicted effectively from genomewide marker effects, predicting the VG of a cross remains difficult.