Artificial intelligence-based systematic study on the multidimensional pharmacological activity and molecular mechanism of the active ingredients of Artemisia argyi

To investigate the pharmacological activities and potential mechanisms of action of the active components in Artemisia argyi with artificial intelligence technology, a search was conducted in the HIT, TCMSP, and TCMIO databases, obtaining 199 active components of A. argyi. A comprehensive set of algorithms, including KNN, MLP, RF, SVM, and models based on Lipinski’s and Veber’s rules, was employed to predict the toxicity and oral bioavailability of A. argyi compounds, identifying 14 components that are non-toxic and have good oral bioavailability. The synthetic accessibility score (SAscore) model was used to analyze the synthetic accessibility of the 14 components mentioned above, and molecular segments were fragmented using BRICS and RECAP algorithms. Mining of the STP and PM databases yielded 406 target proteins for the core components of A. argyi, and Cytoscape was used to screen out 5 core targets: SRC, EGFR, PTPN11, HRAS, and PDGFRB. GO and KEGG enrichment analyses indicated that the core targets were involved in 808 GO enrichment analysis entries and 71 signaling pathways, including EGFR tyrosine kinase inhibitor resistance, gap junction, phospholipase D, and JAK/STAT. Molecular docking results showed that active compounds of A. argyi have a good binding affinity with proteins SRC, EGFR, PTPN11, and HRAS. Cellular experiments have confirmed that ledol, an active component of A. argyi, can promote the proliferation of HUVEC cells within a certain concentration range and can increase the expression of EGFR protein. This study reveals the pharmacological characteristics and potential molecular mechanisms of the active components of A. argyi and lays a solid scientific foundation for its medicinal development.