Machine learning-guided protein selection for functionality-informed cost reduction and quality control in wheat flour dough formulations.

Optimizing wheat flour-protein based dough formulations requires balancing processing performance and cost efficiency. This study evaluated the effects of 14 ingredients - bread flour, eight dietary proteins from both plant and animal sources, three lipids, water, and salt - on wheat flour-protein tortilla dough mechanical properties using Spearman correlation, Random Forest, Elastic Net regularization, Extra Trees regressor-based sensitivity analysis, and cosine similarity analysis. In this study, bread flour was found to be the primary determinant of dough mechanical properties, with greater influence on textural than rheological properties, likely due to the dominant role of starch-water interactions in texture development. Proteins, by contrast, contributed more substantially to rheological than textural properties, either by directly modifying the gluten network or by competitively absorbing water during mixing, thereby limiting gluten hydration and development. Among dietary proteins, plant-based proteins - particularly soy, pea, and pea-quinoa - exhibited substantial impacts, whereas animal-based proteins showed comparatively limited influence. Furthermore, cosine similarity analysis revealed that whey and egg white proteins exhibited comparable mechanical effects, as did soy and pea proteins, supporting whey protein as a functionally equivalent and more economical substitute for egg white protein, and soy protein as a viable alternative to pea protein. Notably, whey and egg white proteins displayed opposing functional directions relative to soy and pea proteins, enabling strategic co-incorporation in plant protein-enriched formulations to enhance nutritional value while maintaining mechanical properties. These findings provide a framework for functionality-informed cost reduction and targeted formulation design in wheat flour-protein dough systems.