Evaluation of binding interaction of coumarin antifungals to bilipid membrane using Dock scoring function and Levenberg-Marquardt neural network

Abstract

Recently, investigation on natural antifungal resources has been increased due to vital need for brand new fungicidal agents. Neural network programs have lots of special features that make them suitable for handling complex problems like analyzing different properties of candidate compounds in computer-aided drug design. In the present study, Levenberg-Marquardt neural network (the fastest of the training algorithms) was used, and the relationship between some important thermodynamic and physico-chemical properties of coumarin compounds and their biological activities (measured by binding interaction energy to bilayer membrane of eukaryote cells) were evaluated. A set of already reported antifungal bioactive coumarin and some well known physical descriptors were selected and, by using Levenberg–Marquardt training algorithm, the best architecture of neural model was designed for predicting the effects of new bioactive compounds. Results revealed that the best architecture according to the term of calculation cycles and considering the correlating behaviour and output cycles of calculation was 19-7-6-1. In addition, the results revealed that the most sensitive input are Log P and molar refractivity. Descriptors, viz., surface tension, energy of LUMO and energy of HOMO were the most important inputs. The correlation coefficient between the observed and the interaction energy values was 0.9132. The study also showed that Dock scoring function can be used for modeling of coumarins antifungal bioactivity.