REPOSITORY > RESULTS

Abstract

Globally, there is increasing awareness of the need for sustainable food systems, driven by projected population growth, rising demand for protein, and tightening constraints on land and water resources. Despite high agricultural productivity, the European Union remains highly dependent on imports of protein-rich feed, with approximately 66% of high-protein feed used in livestock production sourced externally. Simultaneously, water scarcity is emerging as a critical issue in southern Europe. Even in a relatively water-rich, Alpine-influenced country such as Slovenia, regional and seasonal water deficits are increasingly common due to increased drought frequency and variability in water availability, which can threaten agricultural productivity. Both of these challenges, although distinct, have driven interest in circular food production systems that reduce waste, optimise resource use, and close resource loops.
Addressing those challenges, this doctoral thesis investigates two sustainable food sources: farmed edible insects and tomatoes irrigated with treated wastewater (TWW), with a focus on their quality. The research covers four main areas: (i) nutritional characterisation of edible insects, (ii) the impact of fermentation on insect quality, (iii) the effects of TWW and the presence of contaminants of emerging concern (CECs) on tomato quality in hydroponic and soil cultivation systems, and (iv) the application of explainable machine learning for identifying the compounds that contribute most to the differentiation of tomatoes across treatments.
This thesis's findings confirmed that farmed edible insects (Schistocerca gregaria, Tenebrio molitor, Acheta domesticus, Locusta migratoria, Gryllus assimilis, Alphitobius diaperinus, and Blaberus discoidalis) are a rich source of essential fatty acids, amino acids, and minerals, with species, developmental stage, and sex significantly influencing fatty acid and polyphenol composition, while amino acid profiles remained consistent between sexes. Fermentation effects were found to be species- and culture-dependent and generally less beneficial than expected, with limited impact on proximate and fatty acid composition but resulting in reductions in essential amino acids (e.g., lysine, methionine, threonine) and oxidation stability (Tenebrio molitor and Locusta migratoria).
The impact of CECs on tomato quality was influenced by their uptake, the cultivation system, and the growth media. In a hydroponic system, higher concentrations of bisphenol S, 2,4-bisphenol F, and naproxen correlated with increased levels of Mn, Cu, and Zn, suggesting modifications in elemental composition. In contrast, lower uptake in soil-grown tomatoes led to minimal differences among TWW treatments, although TWW irrigation resulted in higher B and Na concentrations than potable water. In a hydroponic system, CEC exposure increased yield without affecting sugars, organic acids, carotenoids, polyphenols, or fatty acids, but it significantly decreased specific amino acids and altered volatile organic compound profiles, likely due to stress, endocrine-disrupting or other toxic effects.
The application of an explainable machine learning approach successfully differentiated between tomatoes grown across five treatments (HPW, HSPW, LPW, LWW, and LSWW), achieving high classification accuracy (75–88%) despite a limited sample size. Furthermore, SHAP (SHapley Additive exPlanations) analysis identified key metabolites and elements at both global and local levels and provided insight into feature contributions, including misclassifications.
Overall, this thesis demonstrates that farmed edible insects and tomatoes irrigated with treated wastewater are promising sustainable food sources. The results contribute to the optimisation of insect rearing and processing and reveal the impact of biological factors and fermentation on their composition and quality. The work is also informative for producers, stakeholders, consumers and regulators regarding the potential effects of TWW reuse on product quality. Finally, it demonstrates the value of explainable machine learning as an effective approach for analysing complex datasets and underscores its broader potential in food and agricultural research.