New article in Current Research in Food Science: Predicting Food Taste with Bound-Driven Optimization

Pagkratis Tagkopoulos, Dimitris Sfondilis, Prof. Ilias Tagkopoulos, and Prof. Tarek Zohdi published Predicting food taste with bound-driven optimization in Current Research in Food Science on 8/22/2026.

About this work: Can you predict what a dish will taste like from its ingredients? This work treats recipes as composite materials, adapting mathematical bounds originally developed in materials science to predict five basic tastes—sweetness, sourness, bitterness, umami, and saltiness—from ingredient composition. Using 70 recipes and 209 ingredient-level taste references with trained-panel measurements, the authors found that simple ingredient mixing consistently underestimated perceived taste: 77% of actual taste values exceeded the predicted upper bound, including 97% for saltiness, 93% for sweetness, and 90% for umami. The study links this gap to processing chemistry such as Maillard reactions, caramelization, evaporative concentration, protein hydrolysis, and nucleotide synergy. Adding eight chemistry-informed proxy features reduced average prediction error from 14.7 to 7.3 across sweetness, sourness, umami, and saltiness, while the resulting 10-feature interpretable model performed comparably to a black-box model using 115 ingredient-level features. The authors then used the model for constrained recipe optimization, demonstrating reformulations of pea soup, chocolate-hazelnut spread, and ketchup toward targeted taste profiles.

Reference: Tagkopoulos, Pagkratis, Dimitris Sfondilis, Ilias Tagkopoulos, and Tarek I. Zohdi. “Predicting food taste with bound-driven optimization.” Current Research in Food Science 13 (2026): 101533. doi: 10.1016/j.crfs.2026.101533.