To accelerate the discovery of novel daphniphyllum alkaloids (DAs), an integrated workflow was developed that synergistically integrates feature-based molecular networking (FBMN), machine learning annotation powered by SIRIUS, and high-speed countercurrent chromatography (HSCCC) directed low-sample-loss separation. FBMN analysis was first performed on three Daphniphyllum species to profile their chemical space, leading to the selection of D. paxianum as the most promising source. Subsequent systematic annotation of MS/MS features using the SIRIUS platform predicted multiple known and potentially novel DAs, providing valuable guidance for targeted isolation. Guided by the FBMN and SIRIUS annotation results, a low-sample-loss separation protocol was subsequently implemented, which combined HSCCC, Sephadex LH-20 gel chromatography, and preparative HPLC. The targeted separation approach, in conjunction with comprehensive spectroscopic analysis (IR, HRESIMS, 1D and 2D NMR) and quantum-mechanical computational methods, enabled the efficient isolation and structural elucidation of 13 DAs from D. paxianum, of which three were identified as new compounds (1~3) with various skeletons, along with ten known analogues (4~13). Compounds 1 and 2 belong to the calyciphylline A-type and daphnilactone B-type DAs, respectively, both featuring a 2-azabicyclo[3.3.1]nonane core, with compound 2 characterized by rare C-2 oxidation. Compound 3 is a rare daphnezomine F-type DA with a 2-azabicyclo[5.2.2]undecane core, enriching the structural diversity of DAs. Evaluation of the antiplasmodial activity using SYBR Green Ⅰ method revealed that compound 4 inhibited the growth of the chloroquine-resistant P. falciparum Dd2 strain with a moderate IC₅₀ of (27.7±3.7) μmol/L. Concurrent biocompatibility assessment demonstrated that compound 4 was devoid of detectable cytotoxicity against A549 and HeLa cell lines (cell viability consistently exceeding 75% at 10 μmol/L) and exhibited negligible hemolytic activity (hemolysis rate <5% at 25 μmol/L). The unity of moderate antiplasmodial efficacy and excellent safety profile highlights compound 4 as a promising candidate for further structural optimization and mechanistic investigation as an antiplasmodial candidate compound. The research demonstrated that the FBMN and SIRIUS dual-guidance strategy, in combination with low-sample-loss separation, constitutes a powerful and efficient approach for the precise discovery of novel and bioactive DAs from complex natural materials.

Key words: Daphniphyllum paxianum, Daphniphyllum alkaloids, feature-based molecular networking (FBMN), high-speed countercurrent chromatography (HSCCC), structural elucidation, antiplasmodial activity