Identification of differentially expressed genes in DLTGenes with P < 0.05 were screened as significant differential genes. The above differentially expressed genes were categorized into up and down-regulated genes using Fold change (FC) as the criterion, and the differentially expressed genes with log2FC > 0.3 were up-regulated genes (up), and the differentially expressed genes with log2FC<-0.3 were down-regulated (down). Finally, 240 up-regulated genes and 241 down-regulated genes were obtained, and their volcano maps are shown in Fig. 1(a), and the 100 genes with the largest and smallest differences were taken to draw the heatmap, as in Fig. 1(b). Figure 1(b) before and after show the group of coronary artery disease patients who did not take DLT and the group who took DLT after 8 weeks, respectively.Subsequently, we performed KEGG analysis of the differential genes of DLT. The results showed iron death as the better enriched pathway, and a review showed that a growing number of studies support the notion that ferroptosis — an iron-dependent form of non-apoptotic cell death that involves the accumulation of lipid hydroperoxides — has a pathophysiological role in the development of cardiovascular diseases12. This indicated that the differential genes that have been identified in DLT were somewhat closely biologically linked to coronary-like diseases. Moreover, the data of DLT were obtained from patients with coronary artery disease. This also proves the reliability of the data. We selected pathways with P < 0.05 to be shown in Supplementary Table S1.Fig. 1Visualization of differential genes in DLT. (a) Volcano map of differentially expressed genes in DLT. (b) Heatmap of differentially expressed genes in DLT.Prediction of the new efficacy of DLTMapping the differential genes of DLT to the CMap database, we obtained 2429 small molecule compounds related to their expression. Based on the literature, the group summarized the pharmacological effects of the top five small molecule compounds with clear pharmacological effects, as shown in Table 1. According to the results, we found that the pharmacological effects of DLT’s analogous compounds were mainly related to anti-inflammatory Neuro-modulating and anti-tumors, and its anti-tumor effects were the most prominent. Therefore, we further focused the disease on NSCLC to carry out the follow-up study.Table 1 Pharmacologic effects of DLT and five small molecule compounds.In vitro validation of DLT against non-small cell lung cancerThe lyophilized powder of DLT was dissolved using DEME containing 10% FBS. 1, 3, 5, 8, 10 and 15 mg/mL of medium containing the drug was configured as therapeutic drug. The survival of the cells after 24 h of drug intervention was calculated. Histograms of cell survival and concentration of drug administered were plotted and analyzed for significance using Graphpad Pism9.5 and the results are shown in Fig. 2. According to the results, it can be concluded that the cell survival rate decreased gradually with the increase of the concentration of DLT, and the IC50 was 6.14 mg/mL. So, we concluded that DLT has the activity against NSCLC.Fig. 2Histogram of A549 cells viability with regard to the concentration of DLT.Gene set enrichment analysis of non-small cell lung cancerWe took the differential expression matrix of the GSE18842 dataset analyzed by the limma package as the object of study and performed GSEA. A total of 20,824 expressed genes were analyzed using GSEA. According to the enrichment analysis results, 103 pathways were significantly enriched (P < 0.05), and the top 30 pathways from the results of the analysis were visualized, as shown in Fig. 3. The vertical coordinates in the graph represent the pathways enriched, and the horizontal coordinates represent the proportion of all genes enriched in each pathway. The size of the bubbles in the graph represents the number of genes enriched in this pathway, and the color of the bubbles represents the significance of the enrichment. Cell cycle, DNA replication, Biosynthesis of amino acid, Osteoclast differentiation, Cytokine-cytokine receptor interaction were the first five pathways. After literature review, all of the above pathways are strongly associated with NSCLC29,30,31,32,33,34. For instance, berberine could significantly inhibit the growth of NSCLC cells through down-regulating DNA replication and repair related proteins RRM1, RRM2, LIG1 and POLE2 in a dose-dependent manner in NSCLC cells29. Yang, Y30. et al. found that HIF-1α via METTL3 regulation of the m6A modification of CDK2AP2 mRNA drives smoking-induced progression of NSCLC through promoting cell proliferation. Similarly, Wang, Z32. et al. choose TGF-β, Activin and TRAIL as the key cytokines and cytokine receptors. In their research, the Camellia. leave. saponinstreatment might promote T cell differentiation and tumor immune response by inhibiting the expression of three TGF phenotypes (TGF-β1, TGF-β2, and TGF-β3), thereby inhibiting tumor angiogenesis and invasion. In addition to that, Vorinostat initiates G1/G2 cell-cycle arrest and disrupts vascular endothelial growth factor (VEGF) signaling in a cell-dependent and dose-dependent manner, resulting in anti-NSCLC effects34.Fig. 3Bubble map of the top 30 pathways from NSCLC gene set analysis results.Identification of key targets for the efficacy of DLT in the treatment of NSCLCRemoving the disease pathway, we compared 481 differential genes of DLT and genes enriched in the top 30 pathways by GSEA. By taking intersections, we obtained a target-pathway relationship, which involves 13 targets, and 14 pathways. Thirteen targets were also labeled in Fig. 1(a). Figure 4 shows the KEGG pathway enrichment results, which mainly contained 14 pathways related to DLT. The genes labeled in the figure are key genes.Fig. 4Identification of the chemical constituents at role in DLTFrom the HERB 2.0 database, 1435 compounds that contain 10 traditional Chinese medicines of DLT were obtained after the de-duplication process. Meanwhile, 286 small molecule chemical compounds corresponding to 13 key targets of DLT associated with the treatment of NSCLC were identified from this database after the de-duplication process. The two groups of compounds were further analyzed to obtain a relationship between 38 compounds and 4 targets.Further, we used the SwissADME database to screen 38 compounds. We downloaded the compounds’ Canonical SMILES from PubChem, entered the compounds’ SMILES in SwissADME “Enter a list of SMILES here”, set the conditions of GI absorption as “High”, Durglikeness as at least two “Yes”, MLOGP < 5. After screening, 24 compounds were obtained, which were regarded as the active ingredients that exerted medicinal effects, as shown in Supplementary Table S2. Finally, we obtained a list of relationships involving 7 herbs, 24 compounds, 4 targets, and 6 pathways, as shown in Supplementary Table S3.We proceeded to visualize the network relationship graphs between DLT, single herbs, active ingredients, targets, and pathways using Cytoscape 3.10.0 software. As in Fig. 5, the shades of colors in the graph were arranged according to the value of degree, and the darker the color indicated that in the network of relationships, the stronger the association. The analysis results showed that Quercetin, Luteolin, Scoparone, Isorhamnetin, Eugenol, Genistein, Coumestrol, Hederagenin, and Succinic Acid were the nine compounds with the strongest associations in the network of relationship, and thus we regarded them as key active ingredients. In addition, CCL2, FEN1, TPI1, and RMI2 are the most critical targets.Therefore, based on these database analyses, we predicted that the anti-NSCLC effects of DLT may mainly exert their efficacy through the key active ingredients components of seven herbs, and it affects targets such as CCL2, FEN1, TPI1, and RMI2 through seven pathways.Fig. 5DLT-Herb-Active Ingredient-Target-Pathway network. The green diamonds are DLT, the light pink circles represent compounds, the orange squares represent targets, the light blue squares on the left represent herbs, and the light purple squares on the right represent pathways.External validation of core targetsWe used GEPIA 2.0 online web analytics to detect the differential expression of key core target genes between NSCLC tissues and normal lung tissues NSCLC includes many types such as adenocarcinoma (LUAD), squamous cell carcinoma (LUSC) and large cell carcinoma. Expression analysis and pathological stage analysis of four genes, CCL2, FEN1, TPI1 and RMI2, in LUAD and LUSC using GEPIA2.0 database, and the results were shown in Figs. 6 and 7. In addition, we analyzed the differential expression of the four genes in the GSE18842 dataset and used the ggplot2 package to draw gene expression box-and-line plots as in Fig. 8. By analyzing Figs. 6 and 8, we found that the four genes were differentially expressed in the same way in the clinical dataset as in the GEPIA 2.0 database. This proves the reliability of the GSE18842 dataset. Second, we found that LUSC was significantly different in four key genes (P < 0.05), but LUAD was significantly different only in FEN1 and RMI2 (P < 0.05). Moreover, by analyzing Fig. 7, the expression level of FEN1 and TPI1 changed significantly with the pathological stage in LUAD.Fig. 6mRNA expression levels of core genes in cancer genome map (TCGA) and genotypic tissue expression (GTEx) databases. Red represents lung cancer tissue, and black represents normal lung tissue. *, P < 0.05. (a) CCL2. (b) FEN1. (c) TPI1. (d) RMI2.Fig. 7Stage diagram of core gene mRNA expression levels and pathological stages in the GEPIA database. (a) LUAD-CCL2. (b) LUSC- CCL2. (c) LUAD-FEN1. (d) LUSC-FEN1. (e) LUAD-TPI1. (f) LUSC- TPI1. (g) LUAD-RMI2. (h) LUSC-RMI2.Fig. 8Expression of core targets in the GSE18842 dataset.Molecular docking analysisBased on the above results, the compounds of the nine core components that are associated with four core targets were selected as docking compounds for simple validation of the above results. Protein PDB structures of CCL2 (P13500, 1DOK), FEN1 (P39748, 5ZOD), TPI1 (P60174, 4POC) and RMI2 (Q9H9A7, 3MXN) were downloaded through the RCSB database. Nine compounds in 3D SDF format files were downloaded by the PubChem database.The results of the docking are shown in Table 2. Kcal/mol is the binding energy obtained by the software. Its 3D model is illustrated in Fig. 9. It is generally accepted that a docking score of less than 0 kcal/mol indicates that the component has the potential for spontaneous binding to the target, and less than − 7 kcal/mol is considered to have strong docking affection35. From the results, it could be concluded that the active compounds all bind well to the active pocket of the protein and all had the potential for spontaneous binding, which proved the accuracy of the above projections results. Hederagenin had the strongest affinity with CCL2. Hederagenin performed interactions with amino acid residues in the CCL2 active site, such as GLU-50, THR-32, and SER-33 through five hydrogen bonds with bond lengths of 1.9, 1.9, 2.8, 2.8, and 1.6, respectively.Based on the docking results, we demonstrate the reliability of the predictive method.Table 2 Ligand-Protein docking results.Fig. 93D interactions of active compounds with protein. (a) quercetin-CCL2. (b) Isorhamnetin-CCL2. (c) eugenol-CCL2. (d) Hederagenin-CCL2. (e) scoparone-CCL2. (f) Luteolin-CCL2. (g) Genistein-CCL2. (h) Coumestrol-FEN1. (i) Genistein-FEN1. (j) Luteolin-FEN1. (k) Coumestrol-RMI2. (l) Succinic Acid-TPI1.Molecular dynamics simulationsHere, we performed dynamical simulations of Hederagenin in complex with the 1DOK protein with the lowest molecular docking binding energy described above to further study the interaction between the active compound and the target protein. After running MD simulations for 100 ns, we analyzed the dynamic behaviors of the complex.The RMSD was used to determine whether the system reached dynamic equilibrium or not, and from the RMSD (Fig. 10a), the ligand-protein complex had been stable in the kinetic simulation system, almost equilibrated near 0.30 nm, which also proved the reliability of the protein-ligand complex docking results and the stability of the complex structure. From the RMSF (Fig. 10b), it could be obtained that residues 17–20, 22–25, 45–50, and 55–60 of the protein are more flexible. Hydrogen bonding as well as hydrophobic interactions play an important role in the preservation of protein conformation. From Fig. 10(d) we saw that hydrogen bonding, as a strong non-covalent force, was stabilized in the complex, which is dominated by hydrogen bonding between some key residues in the proteins and some important groups in the small molecules, with up to five hydrogen bonds in the system. From Fig. 10(c), it can be seen that the ligand acts as the active site of the protein mainly by binding to 15 amino acids, mainly residue 49 of LYS amino acid of A chain, residue 7 of ALA amino acid of B chain, residue 9 of VAL amino acid of B chain and residue 31 of ILE amino acid of B chain.Based on the above results, we further demonstrate the reliability of the prediction method.Fig. 10Molecular dynamics simulation analysis of the Hederagenin-CCL2 complex. (a) RMSD curve of ligand (Hederagenin, red line), protein (CCL2, blue line), and complex (black line). (b) RMSF curve of CCL2(black, chain A; rad, chain B). (c) The energy contribution of hot residue of the complex. (d) Hydrogen bonds of the complex.
