An experimental approach to study multi-species diseases interactions in wheat using machine learning-aided image analysis

Abstract

AbstractFungal plant pathogens causing head blight and leaf blotch diseases are one of the most important threats to cereals such as oat and wheat. Although different resistant varieties have been developed, these diseases are still hard to control thus driving the use of chemical fungicide in Europe and worldwide. Plant breeding programs to develop new varieties with quantitative resistance could result in a longer resistance to the pathogens but require scalable quantitative methods to analyze complex phenotypes. Additionally, in nature, several diseases can occur at the same time due to the coexistence of different pathogen species, thus increasing the genetic complexity of the pathosystem. In the present study we develop a reductionist strategy to study disease resistance at a higher level of organismal complexity, through the application of machine learning to image analysis of artificial pathobiomes. Our results show that such approach enables a meaningful simplification of complex plant multi-pathogen species interactions, allowing the analysis of specific pathogen-pathogen interactions to unravel hidden phenotypic layers that are not visible or quantifiable under field conditions.