Foods

Sensory grading of medium/high temperature Daqu (MHTD) remains subjective. Four mature MHTD samples (S1–S4) spanning practitioner grades were profiled by electronic nose analysis, physicochemical analysis, pH-dependent enzymatic profiling (25 enzymes, pH 4–9), 16S rRNA/ITS sequencing, and HS-SPME-GC-MS, and linked to sensory scores by Pearson correlation and leave-one-sample-out (LOSO) cross-validated regression. Sensory scores ranked S1 > S2 > S3 > S4, and electronic nose profiling separated the two higher-grade samples (S1/S2) from the two lower-grade samples (S3/S4). Traditional physicochemical indices did not form a unified grade gradient, whereas enzymatic and microbial profiles showed clear grade-associated differentiation. The 25 enzymes clustered into three pH-adapted modules; a glycosidase/β-1,3-1,4-glucanase module co-varied as a functional unit and was negatively correlated with acidity. Bacterial communities shifted from Weissella (S1/S4) to Companilactobacillus (S2/S3), while fungal communities shifted from Pichia (S1/S2) to Saccharomycopsis (S3) and a mixed community (S4). Volatile profiles were highly sample-specific. Of 66 indicators, four showed nominal associations with sensory scores at p < 0.05, and none survived false-discovery-rate correction. α-Galactosidase was positively correlated, whereas Lactococcus, Thermomyces, and Staphylococcus were negatively correlated. In leave-one-sample-out prediction, Lactococcus (RMSE = 3.02) and α-galactosidase (3.58) gave the numerically lowest prediction errors. Thermomyces and Staphylococcus were not among the modelled indicators, so the four cannot be ranked by prediction error, and these results remain exploratory given the small sample size. Lactococcus and α-galactosidase are therefore candidate markers that require validation in larger, independent cohorts. Volatile compounds and acidity were unreliable surrogates, and none of the indicators proposed here is ready for practical deployment. Even so, they offer candidate objective descriptors that could complement sensory grading of MHTD.

Stress-related proteins are overexpressed in plants under adverse growth conditions, helping them to survive. The identification and quantification of these proteins are limited in non-model organisms. Furthermore, the high concentration of storage proteins commonly found in seeds increases the difficulty of studying them. Chickpea (Cicer arietinum L.) is one of the most consumed legumes worldwide. Nevertheless, the proteins implicated in stress response have not been explored sufficiently in this organism. Light-emitting diode (LED) light has been used to improve crop quality for a long time; however, the stress-related proteins produced under different light wavelengths remain a major gap. This study aimed to identify how the proteins of the chickpea embryonic axis grown under green and red light at 275 µmol·m⁻²·s⁻¹ are affected. Stress-related differentially expressed proteins (DEPs; |log₂ fold change (FC)| > 1 and Benjamini–Hochberg adjusted p-value < 0.05) under green high (GH) and red high (RH) light were identified using proteomics along with deep learning. Two controls were used, control sun (CS) and control dark (CD). Cysteine was the most enriched amino acid under CD and GH (not significant under RH). Moreover, the FTIR amide I profile of the alkali-extracted protein fraction showed significant differences between treatments only in the intermolecular β-sheet, α-helix and high-frequency β components. Twenty stress-related DEPs were identified among the proteins identified at a false discovery rate (FDR) ≤ 1% at both precursor and protein levels; the highest concentration of them was found under red light, while the lowest was under control sun (CS). Proteins were classified into antioxidant defense, protein folding, chaperones and heat shock proteins (HSPs), dehydrins and late embryogenesis abundant (LEA) proteins, redox homeostasis, and other stress-related proteins. These results suggest that red and green LED light modulate stress-related proteins in the embryonic axis of chickpea sprouts; however, their relevance to nutritional quality remains to be evaluated.

This study aims to elucidate the heat and mass transfer mechanisms, quality changes, and drying state prediction during deep-bed hot air drying of wheat. The experiment was conducted to examine the effects of hot air temperature (50, 60, and 70 °C), bed thickness (10, 20, 30, and 40 cm), and initial moisture content (30.21% and 38.99%, dry basis) on the drying kinetics and quality of wheat. A moisture content prediction model was developed based on a GA–BP neural network. Increasing the hot air temperature significantly shortens the drying time of wheat. However, the hot air temperature of 70 °C caused a decrease in wheat brightness (L value) and pronounced deterioration in wet gluten content, bulk density, and sedimentation value. Hot air temperatures of 50 °C and 60 °C were found to provide the optimal balance between drying efficiency and quality preservation. A vertical moisture gradient was induced by bed thickness, and the grain bed thickness should be controlled at 30 cm or less. The Modified Henderson and Pabis model provided the best fit to the drying data (R² ≥ 0.9995). At the same drying air temperature, the material with an initial moisture content of 38.99% exhibited lower effective moisture diffusivity and higher apparent activation energy (41.071 kJ/mol) than the material with an initial moisture content of 30.21%. The high-moisture-content material exhibited less deterioration in protein and color than the low-moisture-content material. The GA–BP neural network model of moisture content achieved R² values ranging from 0.8915 to 0.9397, showing potential for state monitoring and intelligent control of the drying process. This study provides theoretical support for drying process optimization and equipment design.