Combining Retip Retention Time Prediction with High-Resolution Mass Spectrometry: A Systematic Analysis of <i>Schisandra chinensis</i> - <i>Evodia</i> Conducted for the First Time

<i>Schisandra chinensis</i>-<i>Evodia</i> Herbal Pair (SEHP) is one of the classic Chinese herbal formulas for treating Alzheimer's disease (AD), but its complex chemical composition renders traditional analytical methods inefficient. Retention time (RT) provides complementary information to mass spectrometry, supporting qualitative identification. To enhance identification accuracy and efficiency, this study pioneered the integration of Retip retention time prediction with five machine learning models (Random Forest, BRNN, XGBoost, LightGBM, and Keras) for the systematic identification of SEHP chemical constituents. Using UPLC-Q-Exactive Orbitrap MS technology combined with MS-DIAL and Compound Discoverer software, 165 compounds were identified, including alkaloids, organic acids, and lignans. The LightGBM model achieved high-precision RT prediction within a ±1 min tolerance, significantly reducing false positive identification rates. Through in vivo experiments, 56 parent compounds and 281 metabolites were identified in plasma, urine, feces, liver, and brain tissues of 3xTg-AD mice, revealing their Phase I and II metabolic characteristics. Network pharmacology and molecular docking analysis suggested that SEHP may exert anti-AD effects by regulating key targets such as TNF, AKT1, STAT3, and inflammation-related pathways, including PI3K and MAPK. This study established an efficient and reliable strategy for identifying Chinese herbal medicine components and analyzing their in vivo metabolism, providing scientific evidence for the pharmacodynamic basis and mechanism of action of SEHP.