Neural network modelling of antifungal activity of a series of oxazole derivatives based on in silico pharmacokinetic parameters

In the present paper, the antifungal activity of a series of benzoxazole and oxazolo[4,5-b]pyridine derivatives was evaluated against Candida albicans by using quantitative structure–activity relationships chemometric methodology with artificial neural network (ANN) regression approach. In vitro antifungal activity of the tested compounds was presented by minimum inhibitory concentration expressed as log(1/cMIC). In silico pharmacokinetic parameters related to absorption, distribution, metabolism and excretion (ADME) were calculated for all studied compounds by using PreADMET software. A feedforward back-propagation ANN with gradient descent learning algorithm was applied for modelling of the relationship between ADME descriptors (blood-brain barrier penetration, plasma protein binding, Madin-Darby cell permeability and Caco-2 cell permeability) and experimental log(1/cMIC) values. A 4-6-1 ANN was developed with the optimum momentum and learning rates of 0.3 and 0.05, respectively. An excellent correlation between experimental antifungal activity and values predicted by the ANN was obtained with a correlation coefficient of 0.9536.