Epistatic Models Improve Prediction of Performance in Corn

Abstract

To be useful, adding epistasis to a prediction model must increase predictive power. The objectives of this study were to determine i) using partial least squares (PLS) techniques, whether the ability to predict performance can be increased by including epistasis in a prediction model; ii) whether relaxing the probability of preselecting a marker or interaction to include in the PLS analysis from 0.001 to 0.01 to 0.05 would increase predictive power; and iii) whether the proportion of variability accounted for could be raised to a level useful in breeding. Data for protein, oil, starch, and grain yield were obtained from 500 S2 lines from the crosses of Illinois High Oil × Illinois Low Oil and of Illinois High Protein × Illinois Low Protein corn (Zea mays L.) strains. Lines per se and testcrosses were evaluated for oil, protein, and starch, and only testcrosses for grain yield. Adding epistasis to a model significantly increased predictive power, as did increasing the probability level for inclusion of significant markers and epistatic effects in the PLS analysis from 0.001 to 0.01 to 0.05. With epistasis in the model and P = 0.05, correlations of predicted and observed means were high enough to suggest they could be useful in breeding.