Establishing forensic food models for authentication and quantification of porcine adulterant in gelatine and marshmallow

Abstract This study aims at authenticating gelatine sources using the incorporation of amino acid (AA) analysis via Ultra-High-Performance Liquid Chromatography Diode-Array Detector (UHPLC-DAD) with forensic food models such as discriminant analysis (DA) and principal component analysis (PCA). The multiple linear regression (MLR), principal component regression (PCR) and partial least square regression (PLSR) were compared to select the best model to (1) quantify the porcine adulterant in non-porcine gelatines and (2) quantify porcine adulterant in fish and bovine gelatine. The method linearity was 37.5–1000 pmol/μL with a determination coefficient (R2) of 0.96–1.00 and a total recovery of 85%–111%. The training, testing and validation datasets were established by the AA of fish, bovine and porcine skin gelatines. The DA had successfully classified 100% fish, bovine and porcine gelatines while the PCA identified dominant AA in each gelatine sources. The discriminating model of non-porcine and porcine gelatines (NPPDM) and discriminating model of fish, bovine and porcine gelatines (FBPDM) had 100% correctly classified (1) non-porcine and porcine gelatines and (2) fish, bovine and porcine gelatines. Using the fish and bovine gelatin (PFBG) training dataset, the PCR model was ineffective in quantifying the porcine adulterant in non-porcine gelatines. The MLR was a better model than PCR to quantify porcine adulterant in non-porcine gelatines using porcine adulterant in fish gelatine (PFG) and porcine adulterant in bovine gelatine (PBG) training datasets due to its lower relative error range and average relative error than of the PCR. Likewise, the PCR model of the PFBG training dataset was ineffective to quantify the porcine adulterant, specifically in fish and bovine gelatine. In contrast, the MLR of PFG and PBG training datasets was the best model for quantifying porcine adulterant in fish and bovine gelatines, respectively. The MLR was ineffective to classify porcine gelatine in marshmallow using the PFG training dataset. However, the MLR had successfully quantified porcine gelatine in marshmallow using PFG and PBG training datasets. Since the training, testing and validation datasets were established by the fish, bovine and porcine skin gelatines, the NPPDM, FBPDM and MLR were best applied for these gelatines. This study anticipated that the regulatory bodies might adopt this approach to establish a standard of authentication of gelatine products.