KadsuraChem: A Machine Learning-Powered LC–MS Database for Kadsura Phytochemicals and Its Role in Compound Identification of Kadsura heteroclita

Kadsura plants, rich in bioactive dibenzocyclooctadienes and triterpenoids, are widely used in traditional medicine and the food industry. Previous LC-MS research identified only 25 dibenzocyclooctadienes from K. heteroclita, with issues including isomer confusion and false positives. To overcome these hurdles, this study developed machine learning models based on 161 previously isolated dibenzocyclooctadienes and triterpenoids from Kadsura to predict the retention times (RT) and fragmentation patterns of reference-standard-lacking compounds from Kadsura or Schisandraceae. The KadsuraChem LC-MS database was constructed by integrating RT, fragmentation patterns, MS2 spectra, and chemical information on these constituents using mzVault and MassLists, which effectively minimized false positives and facilitated isomer differentiation, significantly improving the accuracy and efficiency of identification. Additionally, K. heteroclita roots and fruits were found to contain higher levels of lignans and triterpenoids than the stems and leaves, laying a solid foundation for further research and utilization of K. heteroclita.