First, the targets of 18β-GA were predicted in the Swisstargetprediction database and the PharmMapper database, and the drug-target network diagram was drawn using Cytoscape software. Then, the targets of renal cancer were predicted through the GeneCards database. In the GEO database, the keyword “renal cancer” was used for retrieval, and the dataset GSE46699 was selected. The conditions P < 0.05 and |LogFC|≥1 were set for differentially expressed genes, and a volcano plot was drawn using Graphpad. At the same time, the intersection of the results of the three was taken to obtain the intersection target diagram of 18β-GA-renal cancer-GEO.

The GSE46699 dataset was subjected to differential gene analysis through the oebiotech website. Subsequently, the intersection of drug genes, renal cancer genes, GEO differential genes, and WGCNA genes was taken. Then, the corresponding GEO data of the intersection targets were imported into WeishengXin to generate a heatmap. The intersection targets were sorted by LogFC values and a bar chart was drawn in WeishengXin. Meanwhile, principal component analysis was conducted in the PCoA of the Bioladder website.

The targets identified in the prior step were uploaded into the String database, followed by data retrieval and subsequent import into Cytoscape for protein-protein interaction network analysis. Genes with a Degree ≥ 4 were selected as core targets, and the Degree values were input into the Chiplot website for visualization. The GEPIA 2.0 website was used to obtain the correlations among the core targets, and then the intra-group correlation heat map was drawn using Chiplot. The overlapping target genes were uploaded to the DAVID database to perform Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. The top-ranked results of the GO analysis were imported into the Sangerbox website to draw bar charts, and the KEGG results were used to draw enrichment analysis circle diagrams. After sorting by P-value, the Sankey diagram of the pathways was drawn using Microbiomics.

We conducted Lasso, SVM and Random Forest analyses on the intersectional targets obtained from protein-protein interactions using R language to explore more clinically significant targets. Meanwhile, we constructed training and validation sets in Graphpad to verify the analysis results.

The clinical relevance of the core targets was analyzed as follows. Firstly, the Sagnerbox online analysis tool was used to obtain the mRNA expression of the core targets. Then, the targets were input into the GEPIA2.0 website to draw the copy number graph. Next, on the UALCAN website, the tumor type was selected as “Kidney renal clear cell carcinoma” to obtain the box plot of the target protein expression. Subsequently, on the GSCA online website, “Expression” and “Expression & Subtype” were selected to analyze the expression of the targets in different subtypes of renal cancer. Finally, the genes were input into the GEPIA2.0, and the stage graph of renal cancer was drawn in the Expression Analysis.

The core targets were entered into the Human Protein Atlas (HPA) database, where the TISSUE option was set to “KIDNEY” and the PATHOLOGY section was configured to “CANCER” with the subtype specified as “renal cancer,” enabling retrieval of immunohistochemical expression profiles in both normal kidney and renal cancer tissues. Immunofluorescence images of the target proteins were acquired using the SUBCELL module of the HPA database. Additionally, the targets were submitted to UALCAN, selecting the “Kidney renal clear cell carcinoma” dataset to evaluate their survival prognosis.

The core targets were uploaded to the GSCA website, and the tumor type was specified as KIRC. In the “Mutation” module’s “SNV summary” section, the mutation sites and types of SNVs were obtained, and a heat map of the harmful SNV mutation frequency in renal cancer was also obtained. In the “CNV” module, bubble plots depicting heterozygous and homozygous copy number variations of the core targets were generated. Subsequently, the core genes were analyzed using the CAMOIP website to obtain the mutation correlation map and the box plot illustrating microsatellite instability (MSI) expression levels. The occurrence of tumors is related to gene mutations in the body, but the HHR and MMR systems in the body help to repair them. The core genes and the genes of the HHR and MMR repair systems were input into the GEPIA2 website, with “KIRC Tumor” selected. Only the R value data was taken, and the obtained data was imported into the Chiplot website for plotting. A heat map of core genes related to the HRR and MMR repair systems was obtained.

The core target genes were introduced into the UALCAN platform utilizing TCGA dataset for further analysis. The TCGA dataset selected “Kidney renal clear cell carcinoma” and “Methylation” as the results to obtain the gene methylation expression level graph. The TIDE website can analyze the relationship between the methylation level of core targets and survival prognosis. On the TIDE platform, the “Query Gene” function was selected to upload the core targets, and the “CTL Cor” results were retrieved to analyze the association between levels of gene methylation and markers of cytotoxic T lymphocyte (CTL). The “Risk” result was the survival curves of high methylation and low methylation subgroups of genes.

Access the Sangerbox website and select immune infiltration analysis under pan-cancer analysis. Choose TCGA-KIRC in ‘CODE’ to generate scatter plots for the stromal, immune, and estimated scores of the core targets. In the pan-cancer analysis, select immune checkpoint gene analysis and choose ‘KIRC’ in ‘CODE’ to generate the correlation heatmap between core targets and immune checkpoints. Access the TISCH website and choose the ‘KIRC_GSE121636’ dataset to retrieve core gene expression data in immune cells. Subsequently, examine the expression of immune-related cells and core genes in renal cancer. On the TIMER2.0 website, enter the core targets, select “Cancer associated fibroblast, Macrophage, Monocyte, T cell CD8+”, filter out the “KIRC” data and import it into Chiplot to draw the correlation heatmap to show the correlation.

The core targets were uploaded to the TISIDB platform, and the ‘Drug’ option was selected to create the immunotherapy-related network map of the core genes. The CAMOIP website’s “Immune Infiltration” module was used to analyze the associations between core gene expression levels and key immune-related cytokines INF-γ and TGF-β. Lastly, the key genes were input into the “Drug” section of the GSCA database to obtain drug sensitivity data and matched mRNA expression levels from the GDSC and CTRP databases.

Input the core targets respectively into the PDB database, and download the 3D structure files of the core targets. Use Open Babel software to convert the file format to PDB format. Download the SDF structure of 18β-GA from PubChem and analyze it on the CB-Dock2 website. Record the corresponding Vina score values and import the data into ChiPlot to draw a heatmap for display. At the same time, utilize PyMOL software to visualize the molecular docking outcomes.

Next, we conducted toxicity predictions for 18β-GA on the ProTox 3.0 - Prediction Of Toxicity Of Chemicals website and summarized the results.

Using the TF-Target Finder database, we identified the upstream transcription factors of MAP1LC3B. Then, we searched for the downstream targets of MAP1LC3B through the GeneMANIA database, HitPredict database and STRING database, and determined the downstream targets by taking the intersection of the three databases. We analyzed the molecular docking ability of MAP1LC3B and its downstream target proteins through the GRAMM database, and then analyzed the binding energy through the PDBe database. Finally, we visualized the docking results using PyMOL software. The websites related to network pharmacology prediction are shown in Table 1.

To identify differentially expressed miRNAs between renal cancer cells (control group) and 18β-GA-treated renal cancer cells (experimental group), comprehensive transcriptome sequencing was performed with three biological replicates per group (n=3), where all replicate samples were derived from independent cell culture batches. Sequencing results were visualized using volcano plots, bar graphs, and heatmaps to intuitively display differential miRNA expression profiles. Subsequently, functional enrichment analysis was carried out on the target genes associated with these miRNAs. Subsequently, the survival outcomes of miR-27a across various cancers were analyzed using the Kaplan-Meier Plotter platform. The expression levels and prognostic significance of miR-27a and miR-27a-5p in renal cancer were assessed via the UALCAN and ENCORI databases, respectively. Moreover, the association between miR-27a-5p and LC3A, LC3B, and LC3C was assessed. We conducted a correlation analysis between the genes targeted by miR-27a-5p and autophagy-related genes. The Supplementary Materials provided an explanation of the representativeness of the transcriptome samples.

The 786-O renal carcinoma cells were sustained in RPMI-1640 medium, whereas ACHN cells were maintained in MEM medium. Both culture media were enriched with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. The cells were maintained in a humidified incubator at 37 °C with 5% CO₂, and their morphology and proliferation were routinely observed using a microscope. Experiments were typically conducted when cells reached 70–80% confluence. 18β-Glycyrrhetinic acid (18β-GA; purity > 97%, Cat. No. G10105-10G; Sigma, USA) was dissolved in DMSO to create a 10 mM stock solution for later applications. A complete list of experimental materials is provided in Table 2.

Renal carcinoma cells were plated into 96-well plates at a density of 8,000 cells per well, while HK-2 cells were seeded at 6,000 cells per well. After incubation at 37 °C for 24 hours, various concentrations of 18β-GA and 5-fluorouracil (5-Fu) were applied to the wells. The cells were then cultured for additional time periods of 24, 48, or 72 hours. Cell viability was assessed using the CCK-8 assay kit; following the addition of the reagent, plates were incubated at 37 °C for 1 hour. Absorbance values at 450 nm were subsequently recorded with a microplate spectrophotometer. All experiments were performed in triplicate to ensure reproducibility.

Well-conditioned renal cancer cells were digested and counted, and the cell density was set to 4×10⁴ cells/mL. Subsequently, cells were seeded into six-well plates at a volume of 2 mL culture medium per well and incubated for 24 hours. Once cell confluence reached approximately 30%, the initial medium was discarded and substituted with 1 mL of new medium. Lentiviral vectors encoding GFP-tagged miR-27a-5p knockdown, miR-27a-5p overexpression, or corresponding negative control (empty vector), provided by Shanghai Genechem, were then added. The lentivirus volume was determined according to the multiplicity of infection (MOI) and viral titer using the formula: virus volume = (MOI × cell number)/titer. For infection enhancement, 100 μL of HiTransG A was added to ACHN cells, while 100 μL of HiTransG P was added to 786-O cells. The plates were gently mixed and placed in a 37 °C incubator. After 10 hours of transduction, the medium was changed to 2 mL of fresh complete medium to sustain further culture. Transfection efficiency was monitored at 24, 48, and 72 hours post-transduction under a fluorescence microscope by assessing GFP signal in both bright-field and GFP-specific channels. Once transfection efficiency exceeded 90%, downstream experiments were initiated.

Two cell lines were seeded into 6-well plates at a density of 500 cells per well, with three replicate wells per experimental group. Cells were treated according to their respective experimental conditions and cultured with medium refreshed every three days. After 10 days, the experiment was concluded. Cells were rinsed with PBS, fixed for 30 minutes using 4% paraformaldehyde, and subsequently rinsed again with PBS to eliminate the fixative. Subsequently, 2 mL of 2% crystal violet solution was added to each well for staining at room temperature over 15 minutes. Following two PBS washes, the plates were air-dried before imaging. Colonies with diameters exceeding 1 mm were counted as individual clones. The entire experiment was independently repeated three times.

Well-growing renal cancer cells were adjusted to an appropriate concentration and inoculated into 60 mm culture dishes. After complete adhesion, starvation treatment was performed to bring the cell cycle of each group to the same level. Subsequently, cells were processed based on their assigned experimental conditions. Following a 24-hour incubation period, they were harvested after trypsinization, washed using PBS, and then fixed overnight at 4 °C with 70% precooled ethanol. The fixative was removed by PBS washing, followed by the addition of 500 μL of freshly prepared cell cycle staining solution (Rnase A:PI = 1:9). The cells were kept at room temperature in the dark for a 30-minute incubation period, after which cell cycle distribution was analyzed using flow cytometry. This procedure was performed in triplicate to ensure reproducibility.

Well-growing renal cancer cells were inoculated in 6-well plates. Subsequently, the cells were subjected to treatments as defined by their respective experimental groups. On the following day, the cell supernatant and cells were harvested, washed twice with PBS, and the supernatant was discarded. Then, 500 μL of Binding Buffer was added and thoroughly mixed. After adding the staining solution and ensuring complete mixing, the samples were kept at room temperature in the dark for a 15-minute incubation period. Apoptosis analysis was performed by flow cytometry within one hour. The experiment was conducted in triplicate to ensure consistency and reliability of the results.

Renal cancer cells were collected in a sterile and enzyme-free tube, washed with PBS, and then 1 mL Trizol was added and mixed well by pipetting. After ice bath, 200 μL chloroform was added and mixed well. The mixture was kept on ice for 10 minutes and then centrifuged. The upper layer was transferred to a new tube, and an equal volume of isopropanol was added and gently mixed. After ice bath and centrifugation, the supernatant was discarded, and 70% ethanol was added and mixed well. After centrifugation, the supernatant was discarded and the precipitate was air-dried. The precipitate was dissolved in ddH2O. 1 μL of RNA sample was taken to measure the OD value and concentration. An OD260/OD280 ratio of 1.8-2.1 was considered as qualified purity. cDNA reverse transcription was performed using PrimeScript™ RT reagent Kit with gDNA Eraser. Each group had 3 replicates. Data were analyzed by the 2^-△△CT method.

Protein levels in the samples were determined using a BCA protein assay kit. After separation by SDS-PAGE and transfer to PVDF membranes, proteins were fixed onto the membrane surface. The membranes were subsequently blocked with a rapid blocking solution for 15 minutes, followed by overnight incubation at 4 °C with specific primary antibodies. Following extensive washing, the membranes were incubated with corresponding secondary antibodies at room temperature for 2 hours. The PVDF membranes were immersed in the ECL working solution for 2 minutes and exposed using a chemiluminescence imaging system. The protein bands were quantified for protein expression using Image J.

All quantitative data are presented as mean ± standard deviation, based on at least three independent replicates. Statistical evaluations were conducted using GraphPad Prism 10. Group comparisons were conducted via one-way ANOVA. In cases of homogenous variances, ANOVA was applied; for heterogenous variances, Tamhane’s T2 test was used. A p value below 0.05 was considered statistically significant.

Network pharmacology predicted 375 targets for 18β-GA (Figure 2A). The GEO dataset GSE46699 of renal cancer chips was analyzed, and 1322 up-regulated genes and 1045 down-regulated genes were obtained (Figure 2B), with 6251 renal cancer genes (Figure 2C). Intersecting the up-regulated and down-regulated genes of renal cancer with drug targets yielded 421 and 346 targets, respectively (Figure 2D). Next, the intersection of 18β-GA, renal cancer and differentially expressed genes was taken, and a total of 35 targets were obtained (Figure 2E).

As shown in Supplementary Figure 1A, the power value used in this analysis result is 24. Using the chosen power value, genes arecategorized into 11 modules, excluding the grey module, which lacks reference significance, as depicted in Supplementary Figures 1B, C. Pearson correlation analysis was performed to determine the correlation coefficients andcorresponding p-values between module characteristic genes and traits. Modules associated withtraits were identified using a cutoff of |correlation coefficient| ≥ 0.3 and p < 0.05. As illustrated in Supplementary Figure 1D, the turquoise, black, and grey60 modules exhibiting the highest correlation strengths wereselected for further analysis. The intersection of these genes and the previous intersection genes is taken again to obtain 20 more important targets (Supplementary Figure 1E). The heatmap data indicates that about half of these 20 targets are highly expressed inrenal cancer (Supplementary Figure 1F). Supplementary Figure 1G shows the top ten up-regulated and down-regulated genes. PCA is applied to the intersectiontargets to assess sample representativeness, as illustrated in Supplementary Figure 1H. The distinct separation between the tumor and normal groups suggests significant sample representativeness.

To further explore the key targets of 18β-GA in treating renal cancer, we used Cytoscape software to screen 20 targets. As shown in Supplementary Figure 2A, we identified STAT1, KIT, HMOX1, CASP1, HCK, FABP1 and IDO1 as key targets. Supplementary Figure 2B illustrates the use of the Degree value to assess each target’s contribution within the topological map. We continued to analyze their correlations, as shown in Supplementary Figure 2C. Most of the targets showed positive correlations, while FABP1 and KIT tended to show negative correlations. GO analysis indicated that the 20 targets were enriched in BP terms associated with inflammatory responses, CC terms suggesting localization in the cytoplasm and extracellular space, and MF terms linked to functions like protein tyrosine kinase activity (Supplementary Figure 2D). KEGG analysis found that these 20 targets were enriched in three signaling pathways, including “Pathways in cancer”, “PPAR signaling pathway” and “Aldosterone-regulated sodium reabsorption” (Supplementary Figure 2E).

We conducted Lasso analysis on the 7 key targets obtained from the previous step and found that all of them were significant under this algorithm (Figures 3A, B). As shown in Figure 3C, the SVM algorithm also indicated that these targets were significant. As depicted in Figure 3D, the random forest algorithm revealed that the accuracy reached its peak when 6 genes were included. Subsequently, we took the intersection of the genes obtained from the three algorithms and identified FABP1, HMOX1, KIT, HCK, CASP1, and IDO1 as the core targets, as illustrated in Figure 3E. Next, we verified the results obtained from the machine learning algorithms. We found that these targets exhibited significant expression differences in the dataset GSE46699. Using the GSE66272 dataset as the validation set revealed significant differences in these targets between the tumor and normal groups. Therefore, we concluded that the genes obtained through the machine learning algorithms were reliable, as shown in Figures 3F, G. Additionally, we observed that the genes HMOX1, HCK, CASP1, and IDO1 were all related to autophagy. Thus, we hypothesized that there might be a correlation between drug treatment for renal cancer and autophagy. We chose these four genes and the autophagy-related genes MAP1LC3A (LC3I) and MAP1LC3B (LC3II) as primary targets to explor.

Core target expression was examined at both transcriptional and translational levels. At the mRNA level, MAP1LC3A showed high expression in normal tissues, whereas the other targets were predominantly expressed in renal cancer tissues, with MAP1LC3B showed no significant difference in expression between normal and renal cancer tissues (Figure 4A). In copy number analysis, HMOX1, HCK, CASP1, and IDO1 were significantly elevated in renal cancer patients (Figure 4B). At the protein level, HMOX1, HCK, CASP1, and IDO1 were elevated in renal cancer tissues, while MAP1LC3A and MAP1LC3B were highly expressed in the normal group (Figure 4C). Renal cancer staging indicated that MAP1LC3A was highly expressed in stage 2 renal cancer and HCK was highly expressed in stage 3 (Figure 4D). Analysis of core targets in renal cancer clinical staging revealed that MAP1LC3B and HMOX1 levels were elevated in the early stage, whereas IDO1 significantly increased in the middle and late stages (Figure 4E).

Immunohistochemical analysis revealed that MAP1LC3B exhibited high expression in normal tissues, whereas other genes showed elevated expression in renal cancer tissues (Supplementary Figure 3A). Immunofluorescence localization showed that MAP1LC3A and MAP1LC3B were expressed in centriolar satellites and primary cilia, HMOX1 was expressed in the Golgi apparatus and plasma membrane, HCK was expressed in vesicles, plasma membrane and cytoplasm, CASP1 was expressed in the cytoplasm and nucleoplasm, and IDO1 was expressed in microtubules and primary cilia (Supplementary Figure 3B). Elevated levels of MAP1LC3B and HMOX1 were significantly associated with better prognosis and higher survival rates in renal cancer patients, whereas other targets showed no significant impact (Supplementary Figure 3C).

The occurrence of tumors is characterized by genomic alterations. Therefore, we analyzed whether the core targets were changed at the genomic level through SNV and CNV mutation analysis. We found that the correlation between these genes and mutations was not strong (Figure 5A). Then, we further analyzed the frequency of harmful mutations in Hub genes. As shown in Figure 5B, the mutation frequency of HMOX1 was the highest. CNV includes heterozygous mutations and homozygous mutations, among which homozygous mutations will induce more severe disease outcomes. As shown in Figure 5C, we found that the main mutation form of these core targets was amplification. As shown in Figure 5D, the high mutation of these genes was most closely related to VHL and PBRM1, and high expression could promote the expression of proto-oncogenes. Microsatellite instability (MSI) is characterized by alterations in the length of microsatellite sequences due to insertions or deletions during DNA replication, typically resulting from defects in mismatch repair (MMR) functions. Figure 5E’s box plot of hub gene MSI expression levels reveals that only MAP1LC3A positively correlates with MSI, suggesting its elevated expression is linked to increased microsatellite instability.

As shown in Figure 6A, after methylation, HCK is highly expressed in thes renal cancer group, whiles the expression of MAP1LC3B shows no significant difference. The other genes are all at low expression levels. Additionally, as shown in Figure 6B, HMOX1 is positively correlated with cytotoxic T cells, while CASP1 is negatively correlated. The other targets have no significant correlation. As shown in Figure 6C, after gene methylation, MAP1LC3A and HMOX1 are negatively correlated with survival risk, meaning that highly methylated genes are unfavorable for patient survival. We examined the association between mutations in core targets and the DNA repair machinery. The vertical axis represents repair system-related genes, and the horizontal axis represents Hub genes. We found that MAP1LC3A shows a significant negative correlation with two repair systems, while MAP1LC3B, HCK, CASP1, and IDO1 show a positive correlation. Among them, the correlation of MAP1LC3B is the strongest (Figures 6D, E).

The tumor microenvironment is composed of various cells, which can both promote and inhibit tumor development, being complex and variable. Through relevant scoring, we can gain multi-dimensional insights and thus more accurately grasp this delicate environment. Among them, StromalScore is used to assess the proportion of stromal components in renal cancer tissues, ImmuneScore reflects the proportion of immune cells, and ESTIMATEScore reveals tumor purity. Supplementary Figure 4A shows that HMOX1, HCK, CASP1 and IDO1 have significant positive correlations with renal cancer scores, while MAP1LC3A and MAP1LC3B have negative correlations. The single-cell analysis of clear cell renal cell carcinoma (KIRC, dataset GSE121636) primarily illustrates the distribution of immune cell subpopulations within the tumor microenvironment and the expression of key targets in these cells, as depicted in Supplementary Figure 4B. MAP1LC3A is highly expressed in CD8T cells, MAP1LC3B and CASP1 are expressed in various immune cells, HMOX1 and HCK are expressed in monocytes/macrophages, and IDO1 is expressed in dendritic cells. Meanwhile, immune checkpoint analysis indicates that the targets show a positive correlation trend with immune checkpoint molecules (Supplementary Figure 4C). Further, five analysis algorithms were used to analyze the correlations of core targets in immune cells, as shown in Supplementary Figure 4D. MAP1LC3A is negatively correlated with these immune cells, while HMOX1, HCK and CASP1 are positively correlated.

As shown in Supplementary Figure 5A, HMOX1 is closely associated with multiple drugs and targets, while other targets have relatively fewer related drugs. There is no information on drug treatment for MAP1LC3A and MAP1LC3B. IFN-γ has antiviral, antitumor, and immunomodulatory effects, and TGF-β can exert tumor-suppressive effects in early-stage tumors. Supplementary Figure 5B illustrates that HCK, CASP1, and IDO1 show a positive correlation with IFN-γ expression, whereas MAP1LC3A exhibits a negative correlation. Additionally, HCK is positively correlated with TGF-β.GDSC and CTRP are used to screen the sensitivity of antitumor drugs. As illustrated in Supplementary Figure 5C, in CTRP and GDSC, MAP1LC3A, MAP1LC3B, HMOX1, and IDO1 are positively correlated with drug sensitivity, while CASP1 and HCK are negatively correlated with drug sensitivity.

Molecular docking simulations are employed to forecast the interaction affinity between drugs and their target molecules, effectively simulating the drug absorption and metabolic processes that occur in the human body. The binding energy of the core target 18β-GA was predicted through molecular docking, as shown in Figures 7A, B, with the binding energy being less than -7 kcal/mol, indicating a good binding ability. The result was visualized as shown in Figure 7C.

As shown in Supplementary Figure 6A, the toxicity grade of 18β-GA is 4, indicating relatively low toxicity. As depicted in Supplementary Figures 6B, C, the toxicity prediction results are presented. The radar plot displays the confidence level of positive toxicity outcomes relative to the mean toxicity of structurally analogous compounds. In each toxicity response, the closer the blue dot is to the center of the circle, the lower the toxicity; conversely, the farther it is, the higher the toxicity. As shown in Supplementary Figure 6D, the toxicity dose, toxicity grade, and affected organs of 18β-GA are summarized. Overall, its toxicity is relatively low, mainly affecting the respiratory system and the heart.

Since the core targets are mostly related to autophagy, and the MAP1LC3B protein, due to its direct anchoring to autophagosomes and strong association with the core steps of autophagy, has become a key indicator for measuring autophagic activity, we chose to study MAP1LC3B. As shown in Figure 8A, the upstream transcription factor of MAP1LC3B was identified as FOXA1; the downstream target proteins were ATG7 and ATG4B (Figures 8B–D). Based on the analysis of binding energy through molecular docking, the interaction energy between MAP1LC3B and ATG7 was calculated to be -13.9 kcal/mol, and with ATG4B was -11.1 kcal/mol. Therefore, ATG7 with the higher binding energy was selected and visualized (as shown in Figure 8E).

The results of the total transcriptome sequencing showed that miR-27a-5p had the most significant difference in renal cancer after 18β-GA intervention, and its expression was down-regulated after drug intervention (Figures 9A–C). Functional enrichment analysis of predicted target genes regulated by miR-27a-5p using GO and KEGG revealed significant involvement in pathways such as cAMP and AMPK signaling (Figure 9D), and the biological processes were related to cell metabolism, etc (Figure 9E). miR-27a exhibited elevated expression levels in renal cancer (Supplementary Figure 7A), which negatively impacted the prognosis and survival outcomes of patients with this condition (Supplementary Figure 7B). The pan-cancer analysis showed that higher expression levels were negatively correlated with patient prognosis and overall survival (Supplementary Figure 7C). miR-27a-5p exhibited elevated expression levels in renal cancer (Supplementary Figure 7D), which positively influenced the prognosis and survival rates of patients with this condition (Supplementary Figure 7E). miR-27a-5p was negatively correlated with autophagy-related proteins LC3A and LC3C, and positively correlated with LC3B (Supplementary Figure 7F). Supplementary Figure 7G indicated that the genes targeted by miR-27a-5p were mostly positively correlated with autophagy-related genes. The Supplementary Materials (Supplementary Figure 8) demonstrated the accuracy and reliability of the sequencing data.