Rapid visual authentication of high-temperature Daqu Baijiu using porphyrin signal amplification and smartphone-based cloud machine learning

• Published in: Spectrochimica acta. Part A, Molecular and biomolecular spectroscopy Vol. 343; p. 126573
• Main Authors: Zhu, Yanmei, Chen, Hengye, Zeng, Xueqing, Jiang, Xue, Su, Yuanyuan, Cang, Yipeng, Long, Wanjun, Lan, Wei, Fu, Haiyan, She, Yuanbin
• Format: Journal Article
• Language: English
• Published: England Elsevier B.V 15.12.2025
• Subjects: 3,3′,5,5′-Tetramethylbenzidine; Algorithms; Cloud Computing; Cloud-based algorithms; Colorimetry - methods; High-temperature Daqu Baijiu; Machine Learning; Metalloporphyrin; Porphyrins - chemistry; Smartphone; Temperature; Visual array sensor; 3,3′,5,5′-Tetramethylbenzidine; Visual array sensor; TPP_Mg; 3-THPP; TMB; RGB; Cloud-based algorithms; (TMPP)FeCl; RF; TPP; High-temperature Daqu Baijiu; TPP_Cu; Metalloporphyrin; TPP_Zn; DD-SIMCA; RFR
• ISSN: 1386-1425; 1873-3557
• DOI: 10.1016/j.saa.2025.126573

[Display omitted] •Detection mechanism involves Zn2+ coordination competition and π–π interactions.•The porphyrin’s pseudo-peroxidase activity significantly amplifies the signal.•Accuracy exceeds 99 % with computer, and 96.83 % with smartphone-cloud algorithms.•Sensor units TPP_Mg/Zn/Cu contribute the most to the identification accuracy.•Pyrazines contribute most to Baijiu authenticity identification accuracy. High-temperature Daqu Baijiu is known for its superior flavor and value. However, the adulteration of high-grade bottled low-grade Baijiu has become a significant concern. To enhance detection portability, a multi-channel visual sensor array was developed using metalloporphyrin pseudo-peroxidase activity to authenticate the same flavor type Baijiu. The competitive coordination between pyrazine nitrogen atoms and metal ions in porphyrins, along with π–π stacking between aromatic and porphyrin rings, inhibits the pseudo-peroxidase activity of metal porphyrins, leading to 3,3′,5,5′-tetramethylbenzidine (TMB) color changes and amplified spectral signals. Combining the colorimetric array sensor with the random forest (RF) algorithm, compared with a single sensor, it can significantly improve the classification accuracy of Baijiu. Among them, sensor points such as TPP_Mg, TPP_Zn, TPP_Cu, and TPP have made outstanding contributions to the identification of Baijiu. In addition, a DD-SIMCA model for identifying adulteration was constructed, and the recognition accuracy rate reached more than 99%. Another, an original smartphone app based on machine learning cloud algorithm was developed for external method verification, and its recognition accuracy rate reached more than 96.83%. Finally, through random forest regression (RFR) analysis combined with the color response of compounds, it was found that Maillard reaction products such as pyrazines, aldehydes and ketones made significant contributions to the identification and classification of Baijiu. The deviation between the actual value and the predicted value of the compound content predicted by this method is less than 0.2473% ± 1.0785%. This study offers a foundation for developing “instrument-free” rapid visual detection methods for high-temperature Daqu Baijiu.