Evaluation of characteristic aroma compounds in enzymatically enhanced beef tallow using machine learning models

This study investigated the enhancement of beef tallow aroma through lipase-alone, flavourzyme protease-alone, and combined enzymatic treatments. By integrating unsupervised Principal component analysis (PCA), supervised partial least-squares–discriminant analysis (PLS-DA), k-nearest neighbor (KNN), support vector machine (SVM), Gaussian naive Bayes (GNB), logistic regression (LR), and regression-based orthogonal partial least squares regression (OPLSR) modeling, a multi-level machine learning framework was established to systematically elucidate the relationships among enzymatic treatment, volatile compound variation, and sensory perception in beef tallow. Lipase treatment significantly increased the formation of ten straight-chain saturated aldehydes and ( E )-2-alkenals, thereby intensifying fatty notes. Combined lipase–protease treatments significantly promoted the production of nine ketones and furans, contributing to stronger “meaty” and “fatty” odors, whereas protease-alone treatment favored sulfur-containing compounds and terpenes that enhanced the “meaty” and “beefy” attributes. The ML models achieved high predictive performance (R values of 0.83–1.0), confirming their effectiveness in classifying aroma characteristics and identifying key aroma drivers. This integrated analytical approach provides insight into the molecular basis of enzymatic aroma enhancement and offer a scientific basis for future flavor optimization.