• agrochemical fungicides;
  • molecular descriptors based on fragments;
  • linear discriminant analysis


BACKGROUND: The increasing resistance of several phytopathogenic fungal species to existing agrochemical fungicides has alarmed the worldwide scientific community. In an attempt to overcome this problem, a discriminant model based on substructural descriptors was developed from a heterogeneous database of compounds for the design of, search for and prediction of agrochemical fungicides.

RESULTS: The discriminant model classifies correctly 81.95% of the fungicides and 81.54% of the inactive compounds in the training series, with an accuracy of 81.72%. In the prediction series, the percentage of correct classification was 80.59 and 85.56% for fungicides and inactive compounds respectively, with an accuracy of 83.44%. Some fragments were extracted and their contributions were calculated. From the fragments that were determined to make positive contributions to the fungicidal activity, new molecules such as pyrrole derivatives were designed and the probabilities of their being fungicides were calculated. These molecules were correctly classified as potential fungicides.

CONCLUSION: The discriminant model based on substructural descriptors provides a promising methodology for the development of molecular patterns to be used in the design of, search for and prediction of agrochemical fungicides of wide spectrum. This constitutes an alternative for the discovery of compounds that are able to decrease crop losses caused by phytopathogenic fungal species. Copyright © 2011 Society of Chemical Industry