Computational optimization of food formulation and extrusion conditions for nutrient density and antioxidant capacity

Summary

This 2026 study presents a generalizable computational framework that integrates ingredient-level nutrient and bioactive databases (curated from 503 ingredient records via large language model-assisted extraction), mechanistic kinetic models, and machine learning prediction to optimise food formulation and extrusion processing conditions. Using extrusion as a case study, the authors found that ingredient sourcing variability typically dominated nutritional outcomes (3–11 times greater effect than extrusion conditions on NRF 9.3), but that formulation-specific optimisation of processing parameters could yield mean improvements of 10% in combined nutrient density and antioxidant capacity scores. The work bridges food engineering and nutritional science, providing a data-driven approach to systematically improve the nutritional quality of processed foods.

Regional applicability

The framework is generalizable and geography-agnostic, applicable to food product development anywhere processed foods are manufactured. The findings are relevant to United Kingdom food industry and regulatory contexts, particularly for designing nutrient-dense processed foods under evolving nutritional standards; however, transferability depends on whether formulations and ingredient sourcing patterns match UK supply chains and consumer preferences.

Key measures

Nutrient Rich Foods index (NRF 9.3), Ferric Reducing Antioxidant Power (FRAP), ingredient bioactive and nutrient concentrations, extrusion temperature, residence time, coefficient of variation within ingredient sources

Outcomes reported

The study developed a computational framework integrating ingredient databases, kinetic models, machine learning, and global optimisation to predict nutrient retention (NRF 9.3 index) and antioxidant capacity (FRAP) during extrusion processing. Formulation-specific optimisation of extrusion temperature and residence time was predicted to improve combined NRF 9.3-FRAP scores by a mean of 10%, with maxima up to 44%.

Topic tags