Automated annotation of complex natural products using a modular fragmentation–based structure assembly (MFSA) strategy

Complex natural products (CNPs) feature polycyclic structures, multiple consecutive chiral centers, and nonrepetitive structural units. Given their complexity, the structural annotation of CNPs remains a major bottleneck. Here, we introduce the modular fragmentation–based structural assembly (MFSA) strategy for the target CNP structural annotation. The MFSA strategy disassembled the structure based on fragmentation patterns, recognized targets via a pseudo-library, and reassembled the structure using characteristic ions and neutral losses. As a proof of concept, we focused on daphnane-type diterpenoids, a kind of specific bioactive CNP from Thymelaeaceae plants. Furthermore, we present a user-friendly application named CNPs-MFSA coded in Python. Using an in-house daphnane library, CNPs-MFSA outperformed SIRIUS, MS-FINDER, and MetFrag in Top-1 annotation accuracy. By applying CNPs-MFSA to 56 Thymelaeaceae plants, 204 high-confidence daphnanes within 822 annotated results, including 105 previously unreported compounds were found. Aconitine, paclitaxel, and obakunone analogs were incorporated as additional target CNP classes to illustrate the extension workflow.