Quantitative structure–activity relationship study of amide mosquito repellents

Abstract

A quantitative structure-activity relationship (QSAR) study on 43 amide repellents was carried out by the heuristic method in order to reveal the correlations between molecular parameters of these amides and their repellency against Aedes aegypti. Sketches and optimizations of molecular structures were achieved by the Gaussian software package. Generation and screening of molecular parameters were accomplished using CODESSA 2.7.10 software. The leave-one-out method was applied for the model validation. The results showed that a four-descriptor QSAR model with r of 0.897 was obtained. The average r values of the training set and test set of the QSAR model were 0.901 and 0.863, respectively, which suggested that the stability and predictability of the model were confirmed. Analysis of the implications of the descriptors that constitute the QSAR model indicated that all the descriptors were related to the charge distribution over the molecule and affect the dipole moment of the repellents.

Topological QSAR Modelling of Carboxamides Repellent Activity to Aedes Aegypti

“…Repellent activity values were retrieved from Wang et al . for a set of 43 carboxamides: they correspond to the logarithm of the minimum effective dosages (i. e. MED), after conversion into μmol/dm 2 of the results published by Oliferenko et al ., which were originally reported in μmol/cm 2 .…”

“…Log MED activity expressed in μmol/dm 2 covers 2.88 log units (from 0.5185 to 3.3979). The explanation of the conversion of the data from μmol/cm 2 to μmol/dm 2 (resulting in a shift of the reactivity by a factor of 100 in the expression of MED values compared to those modelled by Oliferenko) was not justified by Wang and colleagues . This change in the unit of the endpoint is probably an untended mistake of the authors.…”

“…Recently, Wang et al . re‐examined the repellent activity of the 43 carboxamides studied by Oliferenko in a larger training set, and proposed a multiple linear regression developed using CODESSA descriptors selected by Heuristic method.…”

“…In this paper, we propose, multiple linear regression models developed for the same dataset used by Wang and colleagues, but based on topology‐related 2D structural variables calculated by PaDEL descriptors software v. 2.21 in the software QSARINS. The here proposed approach is based on alternative descriptors to the 2D‐3D CODESSA descriptors used by other authors mentioned above . In addition, we selected the best modelling variables following a procedure based on a genetic algorithm (GA), which involved cross validation up to 30 % out and external validation.…”

“…In addition, we also investigated the use of nonlinear machine learning approaches, which can be easily implemented in R language, to improve the quality of predictions compared to MLR generated in this study. These results were also compared with the CODESSA based model proposed by Wang et al …”