Optimization of chemical fungicide combinations targeting the maize fungal pathogen, Bipolaris maydis: a systematic quantitative approach.

Abstract

To control the southern corn leaf blight, a severe disease of maize around the world, a combination of fungicides is often more potent than using individual fungicides. However, the number of possible combinations increases exponentially with the increase of the number of fungicides combined and their concentrations. It is thus extremely challenging to identify effective fungicide combinations by trial and error from all possible combinations. In this paper, a systematic approach based on a support vector machine, a machine learning algorithm, is proposed to searching for the optimal combinations using only a limited number of measurements. The constructed model also incorporates information related to the inhibition rate (IR) and the cost of each composing fungicide into the optimization process. With this method, we show that only around 130 measurements on a coarse grid of concentrations out of thousands of possible experiments are sufficient to reconstruct the response model and to obtain the optimal fungicide combinations. Experimental results demonstrate that the optimized combinations can achieve an IR greater than 90%, while the required concentrations and the cost of individual fungicides are dramatically reduced. We anticipate that this method should be equally effective in the search for optimal combinations of multiple compounds in other diseases.