Characterization and application of cricket anatomical fractions in sustainable food systems

Abstract

The growing interest in sustainable protein sources has led to increased exploration of edible insects, particularly Acheta domesticus. However, scaling their use in food systems requires a deeper understanding of how feed formulation, processing, and ingredient functionality interact to shape performance and product potential. This dissertation aims to address these knowledge gaps by developing an integrated strategy for the farming, processing, and formulation of A. domesticus-derived ingredients divided in 4 phases: Manuscript 1 (Chapter 2): “Diet optimization by Acheta domesticus self-selection of regional by product ingredients under an industrial farming system” shows that stage-specific self-selection diets using 13 Mexican agro-industrial ingredients (including 6 by-products, 2 food-grade waste materials, and 5 primary products) improved biomass yield (+59 %), survival (+35 %), and feed conversion efficiency (+14 %) versus broiler feed. Ingredient preferences varied by stage (fiber-rich corn bran during initiation, carb-rich corn flour during growth, and lipid-dense toasted sesame pasta during reproduction). This work supports stage-targeted strategies for efficient feed development. Manuscript 2 (Chapter 3): “Novel food ingredients: Evaluation of commercial processing conditions on nutritional and technological properties of edible A. domesticus and its derived parts” characterizes flours from three anatomical fractions (legs+antennae [LF], head+torso [HF], whole body [WF]) processed under three thermal regimes (T1 ≈ 115 °C/2 min @10 psi; T2 ≈ 90 °C/2 min; T3 ≈ 90 °C/30 min). Findings show that anatomical fraction, rather than processing temperature, drives differences in composition and functionality, with leg-based flours offering higher protein content (~70%) and techno-functional properties that are comparable (in practical terms) with whole cricket flour preparations suitable for diverse food applications. Manuscript 3 (Chapter 4): “Unveiling the protein profile and techno-functional potential of A. domesticus protein concentrates: A comparative study of different body parts” evaluates hexane-defatted protein concentrates from legs and antennae (LPC), head and torso (HPC), and whole body (WPC). LPC showed the highest protein content (75.2 g/100 g d.w.) and the highest foaming capacity (75.3%). Functional properties were broadly similar across fractions, while digestibility was highest in HPC (85.5%, PDCAAS = 0.86) and lowest in LPC (PDCAAS = 0.73). ATR-FTIR analysis revealed high β-sheet content in all fractions, with LPC showing the highest proportion (89.6%), indicating extensive protein aggregation that may contribute to reduced digestibility. All fractions displayed tropomyosin bands (~33 kDa), indicating potential allergenicity for individuals with crustacean allergies. These results demonstrate that anatomical fractionation can tailor the functional properties of A. domesticus protein concentrates, though improving the digestibility of LPC remains a key challenge for broader food applications. Manuscript 4 (Chapter 5): “Unlocking the potential of insect and plant proteins: Predicting techno-functional properties with machine learning” presents a dataset of 124 samples [66 single ingredients and 58 binary blends (insect–insect, insect–plant, plant–plant)]. Composition data were used to predict five techno-functional properties: water-holding capacity (WHC), oil-holding capacity (OHC), foaming capacity, oil-in-water emulsifying capacity (O/W EC), and water-in-oil emulsifying capacity (W/O EC). Six machine learning algorithms were tested in multi-output and single-output configurations. Spline regression performed best in multi-output mode (R² = 0.763, MAE ≈ 0.23) and was the most accurate for O/W EC and OHC, while Extreme Gradient Boosting outperformed other models for foaming capacity. Validation with new insect–plant blends showed relative errors within ±1–7% in balanced formulations. The models, while limited by dataset size and missing variables, provide a practical tool for predicting functional properties and reducing empirical trials in ingredient formulation. This dissertation presents an integrated strategy to advance the use of Acheta domesticus as a functional food ingredient. It shows that stage-specific feeding with regional agro-industrial by-products improves farming outcomes, anatomical fractionation creates ingredients with distinct nutritional and functional properties, and machine learning models can predict techno-functional behavior from composition data. Together, these findings support the efficient incorporation of A. domesticus-based ingredients into food systems.