Artificial intelligence and machine learning for texture and sensory optimization in plant-based meat analogs: A review

Abstract

In the production of plant-based meat analogs, mimicking the characteristic texture (toughness, elasticity, fibrous structure) and sensory properties of real meat is a critical technological challenge in food science. Traditional formulation development and extrusion processes are often based on time-consuming, resource-intensive, and costly trial-and-error methods. The aim of this systematic review is to evaluate the optimization potential of artificial intelligence and machine learning algorithms in plant-based meat design and to highlight the advantages they offer compared to traditional methods (e.g., Response Surface Methodology). A review of the current literature shows that machine learning models successfully model the complex and nonlinear relationships between input components and the textural quality of the final product. Neural networks such as decision tree-based algorithms (Random Forest, XGBoost) and Multilayer Perceptrons have been found to provide very high accuracy in texture and rheology prediction. In particular, Multi-Objective Bayesian Optimization has proven to significantly reduce the number of experimental trials required to achieve the target tissue profile compared to traditional methods, maximizing process efficiency. Furthermore, the integration of these algorithms with non-destructive "Green Analytics" tools such as hyperspectral imaging and electronic nose (e-nose) has been shown to support environmental sustainability by reducing sample waste and chemical waste by up to 90%. Consequently, the integration of artificial intelligence and machine learning goes beyond traditional input-to-output approaches in plant-based meat production, enabling the "Inverse Design" vision where the targeted tissue is rationally constructed directly using algorithms. This allows for the much faster development of innovative food formulations that increase resource efficiency and meet consumer expectations.