SPSNP-Smart Prediction System for Natural Product Structures
SPSNP is based on AI and biogenetic building blocks, which represents an innovative paradigm for natural product discovery research. The workflow of this system includes deep learning of building block combination rules, AI-based structure prediction, and screening of predicted structures. The utility of SPMPS was demonstrated by the prediction of phloroglucinol derivatives in the plant Lophostemon confertus.
Introduction of SPSNP
The SPSNP consists of three modules:
- The biosynthesis DL module, in which the chemical structures and corresponding building blocks are used as the training set for the AI program to learn the underlying biosynthetic rules.
- The virtual NP prediction module, in which information from the literature or preliminary experiments is used to determine the building blocks present in a target plant and the abovementioned AI program is used to predict the NPs that these building blocks may generate (called virtual NPs).
- The predicted NP screening module, in which experimental LC-MS/MS data for a target plant extract are compared with the predicted MS/MS data of the virtual NPs and compounds that are present in the target plant (called predicted NPs) are screened out.