Neglecting genotype x environment interaction results in biased predictions from selection index calculations

Abstract

This study investigated effects of neglecting genotype × environment (G × E) interaction on the efficiency of index selection, genetic gain and dollar response. Two indices were established. Index P (P for practical) represents an index that would be used in practice. It assumed that no effects of G × E interactions were present. Index T (T for true) accounted for a prevailing level of G × E interactions in a real data set. Since the true underlying genetic parameters of the data set were unknown, different instances of possible G × E interaction levels for Index T were generated through a sampling procedure. The average difference between the genetic correlations used in Index P and T reduced the efficiency of Index P by between 1% and 24.8%. The total dollar response per year was on average 33% overpredicted, but this could be as high as 114%. Larger differences between the correlations resulted in higher variation in efficiency. Larger G × E interactions resulted in larger overpredictions in total dollar response. The effects of not accommodating or assuming the wrong extent of G × E interaction can be considerable in index calculations.