Non-halal gelatin prediction: a comparative machine learning analysis between OPLS–DA and ANN models

Gelatin is derived from animal collagen, sourced primarily from bovine or porcine, and finds widespread application within the food industry. These issues raise concern over its halal status, particularly among Muslims and Jews, as they adhere to dietary laws prohibiting the consumption of pork and its derivatives. Conventional methods like quantitative Polymerase Chain Reaction (qPCR) and liquid chromatography–mass spectrometry (LC–MS) have limitations due to the deoxyribonucleic acid (DNA)’s reliability and the gelatin's complex composition, respectively. Therefore, this study aimed to explore the application of artificial intelligence (AI)–based machine learning, focusing on amino acid composition for non-halal gelatin prediction. A set of 3,780 data points enabled the analysis of the chromatographic peak areas of 18 amino acids in 210 gelatin samples. Orthogonal partial least squares discriminant analysis (OPLS–DA) and artificial neural network (ANN) compared their performance in machine learning models. The ANN employed resilient backpropagation algorithms that demonstrated high accuracy (98.5%) and regression (R2) of 0.913, with a slightly higher Root Mean Square Error (RMSE) of 0.244. However, OPLSDA demonstrated the best accuracy (100%), R2 of 0.997, and lower RMSE (0.130) compared to the ANN model. The ANN's robustness against outliers and direct output results provided practical advantages, while OPLS–DA offered comprehensive insights and robust discrimination. This study demonstrates the potential of AI-based machine learning in non-halal gelatin prediction, with both models showing the same capability. These findings can be integrated with existing analytical methods to complement the halal analysis, thus ensuring product integrity and upholding halal sanctity