Abstract
Abstract. The purpose of the research is to study the possibility of predicting the fungicidal activity of a large array of organic compounds of different classes in relation to Fusarium oxysporum. Methods. The research is based on the QSAR methodology. The logarithmic form of the minimum concentration of the drug, which inhibits the visible growth of the pathogen, lgMIC, was used as an indicator of fungicidal activity. To construct predictive models, 515 compounds were selected, data on the inhibitory activity of which were obtained from the ChEMBL site. The structure of the compounds is described by molecular structure descriptors calculated by the Dragon 7 computer program; a total of 1600 descriptors are used. Calculations were performed using the PROGROC computer program (PROGgram RObustness Calculation). The algorithm of the program allows using the number of descriptors in excess of the number of substances without preliminary selection. Results. Several models were built with the size of the control sample in the range of 51–61% of the total number of compounds with the following statistical indicators: correlation coefficient R = 0,95…0,978 and standard deviation s = 0,17…0,25. When checking the validity of the models using cross-validation, the following indicators were obtained: R = 0,923 and s = 0,29, the average absolute error is 0,24.The scientific novelty of the study lies in the fact that for the first time the prediction of the fungicidal activity of a large array of organic compounds of different classes against Fusarium oxysporum was performed with high statistical indicators and with an assessment of the predictive power of the models by leave-one-out cross-checking. In the cross-validation mode, lgMIC calculations were also performed for fungicides, the experimental data on the toxicity of which with respect to Fusarium oxysporum are absent.