Bioactive films with essential oils and machine learning for controlling Aspergillus niger growth and fumonisin B₂...

Aspergillus niger is an important species in the fungal community of many foods and is one of the most significant microorganisms used in biotechnology. Some A. niger strains are capable of producing fumonisin B₂ (FB₂) under certain conditions, but little is known about methods to control them, especially in organic foods. FB₂ is a mycotoxin with high toxicity for humans and animals. Effective strategies to control A. niger and FB₂ in agricultural commodities are needed. In this study, ethylene-vinyl alcohol (EVOH) copolymer films containing Origanum vulgare (ORE), Cinnamomum zeylanicum (CIN) essential oils (EO) or individual EO components such as carvacrol (CAR) cinnamaldehyde (CINHO), citral (CIT), isoeugenol (IEG) and linalool (LIN) were evaluated as antifungal and anti-fumonisin agents against two isolates of A. niger isolated from organic maize grains grown in Spain. Radial colony growth rate (GR), effective doses to reduce or completely inhibit fungal growth by 50, 90, and 100 % (ED₅₀, ED₉₀, and ED₁₀₀), and FB₂ levels were determined in all cultures. Assays were performed under different water activities (a_w) (0.95-0.99) and temperatures (20-35 °C). Machine learning (ML) models to predict GR and FB₂ production were developed. Fungal isolate, film type, doses, a_w and temperature had significant effects on film efficacy to control A. niger growth and FB₂ production. Optimal GR was found at 0.99 a_w and 35 °C, and optimal FB₂ production occurred at 0.97 a_w and 28 °C. The most effective bioactive film against A. niger and FB₂ production was EVOH-CINHO, followed by EVOH-IEG and EVOH-CAR; the worst film was EVOH-LIN. A. niger growth was inhibited in the treatments with EVOH-CINHO films at a dose of 250 μg/Petri dish incubated at 20 °C and 0.95 a_w and in the treatments with EVOH films supplemented with IEG, CIT, CAR, or LIN at a dose of 1000 μg/dish. Overall, the XGBoost (extreme gradient boosted tree) and support vector machine (SVM) algorithms generated the most accurate models for predicting GR. The former was also the most effective at predicting FB₂ production. This is the first study on the effect of bioactive films supplemented with EO or its main components on A. niger growth and FB₂ production. The EVOH films incorporating CINHO, IEG, or CIT examined in this study are strongly recommended as biodegradable component in packaging material to control A. niger and FB₂ production in fresh foods, including organic foods.