Based on network pharmacology and machine learning, to explore the molecular mechanism of Xiaoshi Jiangzhi recipe in treating nonalcoholic fatty liver disease (NAFLD). 

The active components of Xiaoshi Jiangzhi Prescription were screened by databases such as TCMSP, HERB, Pubchem and SwissTargetPrediction, and the related targets of NAFLD were obtained by combining Gene Cards and OMIM, and the protein Interaction Network (PPI) was constructed to screen the core targets. GEO data sets were used to analyze the differences of core gene expression and the infiltration characteristics of immune cells, and machine learning model was used to screen characteristic genes and construct the Nomo map. The binding ability of active components to key targets was verified by molecular docking.

There were 184 targets at the intersection of Xiaoshi Jiangzhi recipe and NAFLD. PPI network analysis identified 44 core genes, among which 12 genes such as AKT1, PTEN, APP, SPP1 and ACE were significantly differentially expressed in NAFLD liver tissue (P<0.05). Functional enrichment showed that the core genes mainly regulated adhesion spots and FoxO signaling pathways. Immunoinfiltration analysis showed that the content of activated NK cells in liver tissue of NAFLD patients decreased significantly (P<0.05). The AUC values in the training set and verification set of Nomo graph constructed by using the characteristic genes screened by SVM model are 1.00 and 0.889, respectively. Molecular docking showed that the main components were stably bound to the core targets.

Xiaoshi Jiangzhi recipe may inhibit monocyte infiltration and promote the polarization of M2-type macrophages through signal pathways such as adhesion spots, proteolytic regulation and FoxO, and intervene mast cell degranulation, memory T cell migration and NK cell activation, and multi-dimensionally reshape the immune-metabolic microenvironment of liver to play a therapeutic role on NAFLD. Its mechanism may be related to the intervention of core targets such as AKT1, PTEN, APP, SPP1 and ACE.