Study of the herbicidal profile and soil sorption of sulfonylureas by molecular modeling

Abstract

Weeds can cause great damage to agriculture. Therefore, the development of new herbicides is necessary to guarantee the conservation of agricultural crops and, at the same time, serve as an alternative for the control of resistant pests. Derivatives of sulfonylureas are compounds widely used as herbicides in agriculture. A computational technique used in the design of new agrochemicals, as well as in the interpretation of physical-chemical aspects that explain a certain biological response, is the QSAR (Quantitative Structure-Activity Relationships). Among the existing methods, the present work used the MIA-QSAR technique, which consists of the multivariate analysis of images applied in QSAR, to investigate how the variation in the substituents of a congeneric series of molecules previously synthesized and tested against Brassica napus L. (pyrimidyl acyl thiourea derivatives and sulfonylurea derivatives containing 2,6-disubstituted aryl groups) explains the variation in the herbicidal activity data. The model for pyrimidyl acyl thioureas derivatives proved to be predictive by presenting significant validation statistics parameters, such as r2 pred of 0.833 and r2 of 0.829. Three new herbicides were proposed via in silico method with promising characteristics. As for the herbicides containing 2,6-disubstituted aryl groups, the developed models showed a good performance r2 > 0.90, as well as a high coefficient of determination in the cross-validation leave-one-out q2 > 0.76 and, in addition, four compounds were proposed, and three of these showed herbicidal activity greater than those in the literature. Molecular docking studies were also performed to understand the ligand-enzyme interactions responsible for herbicide activity. A SAR study of soil sorption of known and commercially available herbicides (sulfonylureas) was carried out using experimental log KOC values. Multivariate image analysis (MIA) descriptors were employed to build structure property models to classify these compounds according to their soil sorption capacity. Furthermore, a deeper analysis based on MIA contour maps of PLS regression coefficients and variable importance in projection scores was performed to obtain information about the chemical characteristics responsible for log KOC behavior. A multiple linear regression model obtained from selected descriptors demonstrated high predictability (r2 = 0.95, q2 = 0.84 and r2pred = 0.71) and also provided chemical information for log KOC values. It was found that triazine derivatives, rather than pyrimidine, are less prone to leaching, but substituents attached to the sulfonyl group play an important role in modulating log KOC values.