Prediction of bioactive compounds activity against wood contaminant fungi using artificial neural networks

Abstract

Biopesticides based on natural endophytic bacteria to control plant diseases are an ecological alternative to the chemical treatments. Bacillus species produce a wide variety of metabolites with biological activity like iturinic lipopeptides. This work addresses the production of biopesticides based on natural endophytic bacteria, isolated from Quercus suber. Artificial Neural Networks were used to maximize the percentage of inhibition triggered by antifungal activity of bioactive compounds produced by Bacillus amyloliquefaciens. The active compounds, produced in liquid cultures, inhibited the growth of fifteen fungi and exhibited a broader spectrum of antifungal activity against surface contaminant fungi, blue stain fungi and phytopathogenic fungi. A 19-7-6-1 neural network was selected to predict the percentage of inhibition produced by antifungal bioactive compounds. A good match among the observed and predicted values was obtained with the R2 values varying between 0.9965 – 0.9971 and 0.9974 – 0.9989 for training and test sets. The 19-7-6-1 neural network was used to establish the dilution rates that maximize the production of antifungal bioactive compounds, namely 0.25 h-1 for surface contaminant fungi, 0.45 h-1 for blue stain fungi and between 0.30 and 0.40 h-1 for phytopathogenic fungi. Artificial neural networks show great potential in the modelling and optimization of these bioprocesses.