Maximizing multi-trait gain and diversity with Genetic Algorithms

Abstract

Genetic gain followed by loss of diversity is not ideal in breeding programs for several species, and most studies face this problem for single traits. Thus, we propose a selection method based on Genetic Algorithms (GA) to optimize the gains for multi-traits that have a low reduction of status number (NS), which takes into account equal contributions from individuals as a result of practical issues in tree breeding. Real data were used to compare GA with a method based on a branch and bound algorithm (BB) for the single-trait problem. Simulated and real data were used to compare GA with a multi-trait method adapted from Mulamba and Mock (MM) (a genotypic ranking approach) through a range of selected individuals’ portions. The GA reached a similar gain and NS in a shorter processing time than BB. This shows the efficacy of GA in solving combinatorial NP-hard problems. In a selected portion of 1% and 2.5%, the GA had low reduction in the overall gain average and greater NS than the MM. In a selection of 20%, the GA reached the same NS as the base population and a greater gain than MM for the simulated data. The GA selected a lower number of individuals than expected at 10% and 20% selection, which contributed to a more practical breeding program that maintained the gains and without the loss of genetic diversity. Thus, GA proved to be a reliable optimization tool for multi-trait scenarios, and it can be effectively applied in tree breeding.