RESUMO
Fourier transform infrared (FT-IR) spectroscopy is used throughout forensic laboratories for many applications. FT-IR spectroscopy can be useful with ATR accessories in forensic analysis for several reasons. It provides excellent data quality combined with high reproducibility, with minimal user-induced variations and no sample preparation. Spectra from heterogeneous biological systems, including the integumentary system, can be associated with hundreds or thousands of biomolecules. The nail matrix of keratin possesses a complicated structure with captured circulating metabolites whose presence may vary in space and time depending on context and history. We developed a new approach by using machine-learning (ML) tools to leverage the potential and enhance the selectivity of the instrument, create classification models, and provide invaluable information saved in human nails with statistical confidence. Here, we report chemometric analysis of ATR FT-IR spectra for the classification and prediction of long-term alcohol consumption from nail clippings in 63 donors. A partial least squares discriminant analysis (PLS-DA) was used to create a classification model that was validated against an independent data set which resulted in 91% correctly classified spectra. However, when considering the prediction results at the donor level, 100% accuracy was achieved, and all donors were correctly classified. To the best of our knowledge, this proof-of-concept study demonstrates for the first time the ability of ATR FT-IR spectroscopy to discriminate donors who do not drink alcohol from those who drink alcohol on a regular basis.