Valorization of Spent Millet Flour in Composite Bread: Characterization, Predictive Modeling, and Multi-Objective...

This study valorized spent millet flour (SMF) from traditional pito beverage processing by incorporating it into composite bread formulations. SMF was prepared by oven-drying the pito residue at 60°C, dry-milling, and sieving through a 250 μm mesh to achieve a standardized particle size suitable for bread applications. Response surface methodology (RSM) was employed to develop predictive models and identify optimal substitution levels. SMF exhibited superior nutritional composition compared to wheat flour, with higher crude protein (12.8% vs. 11.5%), crude fiber (8.7% vs. 2.1%), and ash content (2.8% vs. 0.6%), alongside a 33% greater water absorption capacity attributed to fermentation-induced protein modification and fiber concentration. Composite bread was prepared at 10%, 20%, and 30% SMF substitution levels and compared against a 100% wheat flour control bread. Second-order polynomial models developed for twelve quality responses achieved high predictive accuracy (R ² = 0.912-0.988) with non-significant lack-of-fit tests (p > 0.05) and experimental validation errors below 4.5%. Multi-objective desirability function optimization identified 14.2% substitution as globally optimal (D = 0.847), delivering a 9.8% protein increase and 68.4% fiber increase while maintaining consumer acceptability at 7.21/9.0. The reduction in loaf volume with increasing SMF substitution is attributable to a dual mechanism: (i) gluten dilution, whereby replacement of wheat protein with non-gluten millet proteins reduces the viscoelastic network required for gas retention, and (ii) physical disruption of the gas-cell wall structure by coarse insoluble bran particles in SMF, which puncture developing gas cells during proofing and early baking stages. Principal component analysis, hierarchical cluster analysis, and Pearson correlation confirmed the mechanistic trade-off between nutritional enhancement and structural bread quality, with fiber content emerging as the dominant predictor of volume reduction (r = -0.89, p < 0.01). The RSM optimization framework demonstrated here is transferable to other fermented cereal by-product systems across sub-Saharan Africa.