Integer programming as a powerful tool for polyclonal selection in ancient grapevine varieties

Abstract

Integer programming was used as a novel approach for grapevine selection. Several selection criteria were considered using real data to test the method, which was successfully applied to polyclonal selection. Polyclonal selection (selecting a high-performing, balanced mixture of 7 to 20 clones) in ancient grapevine varieties is a selection method that is increasingly used in countries with ancient viticulture. However, to meet the needs of the vine and wine sector, polyclonal selection must take into account several target traits. Polyclonal selection is based on empirical best linear unbiased predictors of genotypic effects obtained by fitting appropriate linear mixed models. This work proposes a multicriteria method for polyclonal selection. A new approach based on integer programming is implemented to perform polyclonal selection considering multiple traits simultaneously. An algorithm that attempts to maximize the genetic gains of selection according to different selection criteria has been developed and tested on real data of important traits obtained in large field trials of four ancient grapevine varieties. Multiple selection criteria were used to perform polyclonal selection of groups of 7 to 20 clones of each variety based on multiple traits. The results showed that integer programming can be useful in polyclonal selection to obtain selected material with high genetic gains in the target traits, while avoiding losses in other equally important traits.