Spectroscopic detection of chemical markers for traceability in complex biological matrices: A case study on wet-blue...

This study investigated whether unique spectral features from tracer compounds could be reliably detected and discriminated from the complex spectral background of biomaterial matrix. Spectroscopy combined with chemical markers present a promising solution for traceability in complex biological matrices, such as leather and other biomaterials involving intricate processing. This study explores the potential of Near InfraRed Spectroscopy (NIRS), both benchtop (350-2500 nm) and miniaturized (900-1700 nm), as a rapid, non-destructive tool for detecting molecular tracers embedded within wet-blue leather. Three candidate tracers-collagen hydrolysate, Cloisite Na⁺ (CNa), and 2-Thiocyanomethylthiobenzothiazole (TCMTB)-were tested across two experimental phases under both industrial tannery and controlled laboratory conditions. Initial analyses using Principal Component Analysis (PCA) indicated TCMTB as the most promising tracer based on spectral distinctiveness. Subsequently, advanced machine learning approaches, including Support Vector Machines (SVM) and feature selection via Genetic Algorithm (GA) coupled with Random Forest (RF), were employed to optimize classification performance and identify key wavelengths contributing to tracer detection. The results confirmed that molecular tracers exhibit identifiable and reproducible spectral signatures, though their detectability is significantly influenced by the inherent variability of the leather matrix and the side of sample scanning. The integration of chemometrics and intelligent feature selection enhances detection sensitivity, highlighting the critical role of spectral range and algorithmic robustness. This study underscores the viability of spectroscopy and machine learning as a platform technology for traceability in complex bio-derived materials, with leather serving as a representative use case.