Assessment of artificial neural networks to predict red colorant production by Talaromyces amestolkiae

Consumer choice is typically influenced by color, leading to an increase in the use of artificial colorants by industry. However, several artificial colorants have been banned due to their harmful effects on human health and the environment, leading to increased interest in colorants from natural sources. Natural colorants can be found in plants, insects, and microorganisms. The importance of evaluating the technical and cost feasibility for the production of natural colorants are important factors for the replacement of artificial counterpart. Therefore, it is highly beneficial to predict the productivity of microbial colorants. The use of statistical methods that generate polynomial models through multiple regressions can provide information of interest about a bioprocess. However, modeling and control of biological processes require complex systems models, because they are nonlinear and non-deterministic systems. In this regard, artificial neural networks are suitable for estimating bioprocess variables with systems modeling. In this work, two different strategies were developed to predict the production of red colorants by Talaromyces amestolkiae, namely simulation by artificial neural networks (ANN) and response surface methodology (RSM). The results showed that the colorant concentration predicted by ANN is closer to the experimental data than that predicted by polynomial models fitted by multiple regression. Thus, this work suggests that the use of ANN can identify the initial conditions of the culture parameters that have the greatest influence on colorant production and can be a tool to be employed to improve the production of biotechnological products, such as microbial colorants.