Detecção e quantificação de cafeína em suplementos de creatina por espectroscopia no infravermelho médio e algoritmos de inteligência artificial
Nenhuma Miniatura disponível
Data
2025-12-03
Autores
Pires, Maria Clara da Cruz
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal do Espírito Santo
Resumo
Due to the economic relevance and widespread popularity of dietary supplements (DS), combined with weaknesses in regulatory oversight, these products have become susceptible to fraudulent practices, particularly adulteration, reinforcing the need for alternative methods to assess their quality and safety. This study aimed to develop machine learning algorithms for the detection and quantification of caffeine (CAF) in creatine (CRE) supplements using mid infrared spectroscopy associated with artificial intelligence (AI) models. Five commercial CRE brands were adulterated with CAF at concentrations ranging from 2 to 20%, generating samples which, together with the corresponding pure formulations, resulted in a total of 56 samples. The mixtures were analyzed in triplicate using a Bruker® ALPHA II FTIR spectrometer (4 cm⁻¹ resolution, 32 scans, 4,000–400 cm⁻¹), yielding 168 spectra, which were processed using Orange Data Mining® (v.3.38.1). The dataset was divided into calibration (CAL, 70%) and prediction (PRED, 30%) sets and assigned to class 0 (n = 18) for pure samples and class 1 (n = 150) for adulterated samples. The following multivariate analyses were applied: Principal Component Analysis (PCA), Support Vector Machine (SVM), and Partial Least Squares (PLS). Two additional blind prediction tests were conducted: one using three known CRE brands (18 adulteration levels, 0–20%) and another using one CRE brand unknown to the models (six adulteration levels), both prepared independently by different researchers. PCA of pure CRE and CAF samples showed an explained variance (EV) of 98.5% (PC1 = 97.4%; PC2 = 1.1%), whereas PCA of the complete dataset showed an EV of 86.75% (PC1 = 68.61%; PC2 = 18.14%). In the test, the SVM achieved sensitivity (SEN) of 100% and specificity (SPEC) of 75%. The PLS yielded a coefficient of determination (R²) of 0.75 and a root mean square error of prediction (RMSEp) of 3.06%. The limits of detection (LoD) and quantification (LoQ) were 0.55% and 1.82%, respectively. In the first blind test, the SVM reached 100% SEN starting at 1.71% adulteration; the PLS achieved an R² of 0.67 and an RMSE of 2.84%. In the second blind test, SEN was 80%, and the PLS achieved an R² of 0.73 and an RMSE of 2.11%. Overall, the results demonstrate that the AI-based algorithms were effective for the detection and quantification of CAF in CRE supplements, representing a practical and scalable tool for assessing product quality and authenticity
Descrição
Palavras-chave
Suplementos dietéticos , Espectroscopia de infravermelho , Alimentos , Inteligência artificial , Spectroscopy Fourier Transform Infrared , Machine learning , Food adulteration , Dietary supplements