Network Pharmacological Analysis of a New Herbal Combination Targeting Hyperlipidemia and Efficacy Validation In Vitro

The network pharmacology (NP) approach is a valuable novel methodology for understanding the complex pharmacological mechanisms of medicinal herbs. In addition, various in silico analysis techniques combined with the NP can improve the understanding of various issues used in natural product research. This study assessed the therapeutic effects of Arum ternata (AT), Poria cocos (PC), and Zingiber officinale (ZO) on hyperlipidemia after network pharmacologic analysis. A protein–protein interaction (PPI) network of forty-one key targets was analyzed to discover core functional clusters of the herbal compounds. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and gene ontology (GO) term enrichment analysis identified significant categories of hypolipidemic mechanisms. The STITCH database indicated a high connection with several statin drugs, deduced by the similarity in targets. AT, PC, and ZO regulated the genes related to the energy metabolism and lipogenesis in HepG2 cells loaded with free fatty acids (FFAs). Furthermore, the mixture of three herbs had a combinational effect. The herbal combination exerted superior efficacy compared to a single herb, particularly in regulating acetyl-CoA carboxylase (ACC) and carnitine palmitoyltransferase 1 (CPT-1). In conclusion, the network pharmacologic approach was used to assess potential targets of the herbal combination for treatment. Experimental data from FFA-induced HepG2 cells suggested that the combination of AT, PC, and ZO might attenuate hyperlipidemia and its associated hepatic steatosis.


Introduction
Hyperlipidemia is a status of an elevated lipid profile in the blood due to lipid metabolic disorders [1]. The WHO announced that hyperlipidemia and high blood pressure, along with alcohol consumption and smoking, are the major causes of fatality in recent years [2]. Hyperlipidemia is generally treated with dietary intervention and hypolipidemic agents by a clinical practitioner according to the lipid profiles of the patient [3]. The clinical features of hyperlipidemia are elevated lipids in the bloodstream, including fat, fatty acids, cholesterol, phospholipids, and triglycerides [4]. The reported environmental factors of hyperlipidemia were obesity, excessive alcohol intake, BMI, and genetic factors, including LDL (low-density lipoprotein), apoA-I, apoE, and microsomal triglyceride transfer protein [5,6]. Dietary patterns fundamentally impact increased LDL cholesterol levels [7]. Elevated LDL cholesterol is a significant clinical marker of patients with atherosclerosis study was performed to verify the mechanism of these herbs estimated by network pharmacology using a fatty acid-induced hepatic steatosis model. Finally, an analysis of the efficacy and mechanism of the three medicinal herbs that can be used in the prescription of hyperlipidemia highlights their potential for treating hyperlipidemia.

Data Preparation Data Acquisition of Herbs from the Online Database
The TCMSP (https://tcmsp-e.com accessed on 12 August 2022) [34], which is a pharmacology database of Traditional Chinese Medicine, was used to collect the compounds and targets of the herbal combination. The oral bioavailability (OB) and drug-likeness (DL) were used for screening of bioactive compounds [18]. OB represents the rate and extent of an active ingredient or active moiety that is absorbed into the blood circulation and becomes available at the site of action [35]. DL can describe the molecular properties affecting the pharmacodynamics and pharmacokinetics [36]. The threshold values of the two indices were ≥30% (OB) and ≥0.18 (DL) [37]. The results from the TCMSP database were reinforced using the OASIS database to sum the additional bioactive compounds with proper literature-based evidence for their activity (https://oasis.kiom.re.kr/accessed on 12 August 2022).

Protein-Protein Interaction (PPI) Network Construction
A PPI network with STRING (Switzerland) was produced using a query list of target genes and exported to Cytoscape software version 3.9.1 (USA), a free software package for visualizing, modeling, and analyzing molecular and genetic interaction networks (confidence score = 0.400) [41]. The total PPI network was clustered functionally using MCODE (USA) and analyzed further with ClueGO (USA) and CluePedia (USA), which are Cytoscape plugins (USA) [41][42][43]. The STRING database aims to provide a critical assessment and integration of protein-protein interactions, including direct interactions [44]. A protein-chemical network was then screened using the STITCH database version 5.0 (USA) by submitting a target gene list [45].

Gene Ontology (GO) Terms and KEGG Pathway Enrichment Analysis
The enrichment of GO terms of BP was analyzed in the DAVID database (USA). Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling-pathway enrichment analysis was performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) v.6.8 (https://david.ncifcrf.gov/, accessed on 29 August 2022) [46]. A list of target genes was submitted, and the gene identifier was set to 'official gene symbol'.
The false discovery rate (FDR) was used as a statistical test method of the enrichment analysis in the DAVID database, which is based on fisher's exact test [47]. The data of relative gene ratio, adjusted p-value, and gene counts of each GO term and KEGG were processed and presented as bubble plots using the R package (ggplot2) and public script in R studio (USA) [48].

Chemicals and Antibodies
Dulbecco's Modified Eagle's Medium (DMEM) was purchased from Hyclone (Logan, UT, USA), and fetal bovine serum (FBS) and penicillin/streptomycin solution were purchased from Invitrogen (Carlsbad, CA, USA). The EZ-Cytox assay kit, obtained from Daeil Lab Service (Chungcheongkuk-do, South Korea), was used to measure cell viability. The phosphorylation-form or non-phosphorylation-form primary antibodies of ACC (Acetyl-CoA carboxylase), AMPK (AMP-activated protein kinase), AKT (Protein kinase B), CPT-1 (Carnitine palmitoyltransferase-1) were purchased from Cell signaling Technology (Berkeley, CA, USA), and β-actin was obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA), which also supplied secondary antibodies. The oligonucleotide primers for real-time qPCR were produced by Macrogen (Seoul, South Korea).

Preparation of Samples
Dried herbs of A. ternata (AT), P. cocos (PC), and Z. officinale (ZO) were purchased from Herbmaul (Chungcheongbuk-do, South Korea). To prepare the extract, dried herbs (100 g) were ground to a powder and extracted with 500 mL in distilled water at 100 • C for 15 min. In addition, the mixed sample was blended with equal weights of AT, PC, and ZO (33 g each). The hot water extracts were filtered twice through 8 µm-pore-size Whatman filter paper, concentrated by rotary evaporation (Buchi, Flawil, Switzerland), and freeze-dried to obtain lyophilized extracts of AT, PC, and ZO. These were then eluted with DPBS and filtered through a 0.22 µm syringe filter before cell treatment.

Cell Culture and Treatment
HepG2 cells (a human hepatocellular carcinoma cell line) were purchased from the Korean Cell Bank (no. 88065, Seoul, South Korea) and cultured in DMEM supplemented with a 1% penicillin/streptomycin and 10% FBS at 37 • C in humidified 5% CO 2 environment. To evaluate anti-intracellular lipid accumulation effects, the FFAs (oleic acid and palmitic acid, 2:1, v/v, respectively) were dissolved in DMSO at 1 mM concentrations. The HepG2 cells were serum-starved for 12 h, cultured with serum-free DMEM containing 1% bovine serum albumin (BSA), and exposed to 1 mM FFAs [49]. Each sample was co-treated during incubation in FFA-BSA complex media for 24-48 h for further analysis.

Cell Viability Assay
The cell viability of HepG2 cells was determined using the EZ-Cytox cell viability assay kit (Seoul, South Korea) according to the manufacturer's instructions with slight modifications [50]. Briefly, HepG2 cells were plated at a density of 4 × 10 4 cells/well in 96-well plates. After 24 h incubation, the medium was changed to FBS-free DMEM containing a serially diluted sample (0-50 µg/mL), treated with different concentrations, and incubated at 37 • C in a humidified containing 5% CO 2 for 24 h. Then, 10 µL of EZ-Cytox reagent was added to each well, and cells were incubated for 1 h. Optical densities (ODs) were measured at 450 nm using a microplate reader (Versamax, Molecular Devices, CA, USA).

Western Blot Analysis
The protein levels related to hepatic steatosis were measured by Western blot. The whole protein was isolated using a RIPA buffer (Thermo Fisher Scientific, Rockford, IL, USA) containing a protease and phosphatase inhibitor cocktail (Gendepot, Barker, TX, USA) after the cells were washed with Dulbecco's phosphate-buffered saline (DPBS). The protein concentrations were determined using a BCA kit (Thermo Fisher Scientific, Rockford, IL, USA). Equal amounts of protein sample (40 µg/mL) were mixed with the 5× Lane Marker Reducing sample buffer (BioRad, Hercules, CA, USA) and denatured at 95 • C for 10 min. The protein lysates were loaded into 7.5% SDS-PAGE gels, electrophoresed, and transferred to PVDF membranes activated by methanol at 100 V for 60 min using an electrophoretic transfer cell (Bio-rad, Hercules, CA, USA). The membranes were blocked with 5% BSA in TBS/T (TBS containing 0.1% Tween 20) for 2 h at room temperature. The blots were incubated with primary antibodies (diluted at 1:1500 in TBS/T containing 3% BSA) overnight at 4°C with gentle shaking. After washing with TBS/T, the membranes were incubated with secondary antibodies (diluted at 1:3000 in TBS/T containing 1% BSA) at room temperature for 3 h. The membrane was detected using a Western blot imaging system (Fusion Solo chemiluminescence system, Vilber Lourmat, Collegien, France), and proteins were visualized using a Super Signal West Pico ECL buffer (Thermo Fisher Scientific, Rockford, IL, USA) [51].

Oil Red O Staining
The lipid accumulation was estimated by staining the intracellular lipid droplets with the Oil Red O reagent (Thermo Fisher Scientific, Rockford, IL, USA). HepG2 cells were plated at a density of 4 × 10 5 cells/well in six-well plates. After 24 h, the cells were incubated with the samples (25, 50 µg/mL) and FFAs (1 mM) for 48 h. The cells were washed with DPBS and then fixed with 10% formalin for one hour at room temperature. After fixation, the cells were washed with 60% isopropanol and stained with a prepared working solution of Oil Red O for 15 min. The stained cells were washed three times with ultrapure water and dried. The cells were examined under an inverted microscope system with the camera (DMI 6000, Leica, Wetzlar, Germany), and the stains were re-dissolved in pure isopropanol to measure OD at 520 nm wavelength [53].

Statistical Analysis
The analysis was conducted using one-way ANOVA in Graph Pad Prism version 5.0 software (Graph Pad, La Jolla, CA, USA). The standard curves were constructed using Excel and PowerPoint (Microsoft, Redmond, WA, USA). The significance of the differences between the untreated controls and the FFs-treated cells, and between the FFA-treated and the sample-treated cells, were determined. The results are presented as means ± SDs, and p-values of <0.05 were considered significant.

Selection of Potential Compounds from AT, PC, and ZO
The data of AT, PC, and ZO compounds and targets were retrieved from TCMSP and OASIS. Table 1 lists the compounds studied, and Figure 1A presents a data plot. ZO, PC, and AT have 6, 7, and 33 compounds; all three herbs commonly share one compound (palmitic acid).
The analysis was conducted using one-way ANOVA in Graph Pad Prism version 5.0 software (Graph Pad, La Jolla, CA, USA). The standard curves were constructed using Excel and PowerPoint (Microsoft, Redmond, WA, USA). The significance of the differences between the untreated controls and the FFs-treated cells, and between the FFA-treated and the sample-treated cells, were determined. The results are presented as means ± SDs, and p-values of <0.05 were considered significant.

Selection of Potential Compounds from AT, PC, and ZO
The data of AT, PC, and ZO compounds and targets were retrieved from TCMSP and OASIS. Table 1 lists the compounds studied, and Figure 1A presents a data plot. ZO, PC, and AT have 6, 7, and 33 compounds; all three herbs commonly share one compound (palmitic acid).

Target Prediction
In total, 168 and 291 targets for hyperlipidemia from disease databases were screened for three herbs from TCMSP ( Figure 1B,C). As a result, 41 common targets assumed to be involved with hyperlipidemia were obtained between the target lists of the disease database and TCMSP ( Figure 1D). The distributions of targets were visualized using a Venn diagram.

PPI Networks Construction and Analysis
The PPI Network was built for 41 targets expected to be significant for hyperlipidemia using the STRING Database. The full network consisted of 41 nodes and can be divided into two clusters using the MCODE of Cytoscape (Figure 2). The targets were arranged in order of importance of network parameters of degree, betweenness centrality, and stress. AKT Serine/Threonine Kinase 1 (AKT1), peroxisome proliferator-activated receptor γ (PPARγ), and prostaglandin-endoperoxide synthase 2 (PTGS2) act as important targets (Table 2).

Target Prediction
In total, 168 and 291 targets for hyperlipidemia from disease databases were sc for three herbs from TCMSP ( Figure 1B,C). As a result, 41 common targets assume involved with hyperlipidemia were obtained between the target lists of the d database and TCMSP ( Figure 1D). The distributions of targets were visualized u Venn diagram.

PPI Networks Construction and Analysis
The PPI Network was built for 41 targets expected to be significa hyperlipidemia using the STRING Database. The full network consisted of 41 nod can be divided into two clusters using the MCODE of Cytoscape (Figure 2). The were arranged in order of importance of network parameters of degree, betwe centrality, and stress. AKT Serine/Threonine Kinase 1 (AKT1), peroxisome prolif activated receptor γ (PPARγ), and prostaglandin-endoperoxide synthase 2 (PTGS2 important targets ( Table 2).

CTP Visualization of Cytoscape
Forty-one targets were analyzed through the KEGG pathway. Based on this, the compounds-targets-pathways network was visualized in Cytoscape ( Figure 3). The top five pathways expected to be effective for hyperlipidemia were identified by importing 41 targets into the DAVID database. Each node was connected to the edges of the compounds, targets, and pathways that are expected to be correlated with each other. Table 3 lists the top enriched KEGG pathways related to lipid metabolism analyzed using the target gene list.

Network Analysis Using ClueGO, CluePedia
Using the full target list of the herbal combination, the network of major pat and its component targets was visualized using ClueGO and CluePedia in Cyt (Figure 4). The ClueGO network could identify and visualize the interaction of pat and their targets. As a result, the pathway of the 'AGE-RAGE signaling pathw diabetic complication' interconnected two other significant pathways in lipid diso such as 'regulation of lipolysis in adipocytes' and 'PPAR signaling pathway', with th

Network Analysis Using ClueGO, CluePedia
Using the full target list of the herbal combination, the network of major pathways and its component targets was visualized using ClueGO and CluePedia in Cytoscape ( Figure 4). The ClueGO network could identify and visualize the interaction of pathways and their targets. As a result, the pathway of the 'AGE-RAGE signaling pathway in diabetic complication' interconnected two other significant pathways in lipid disorders, such as 'regulation of lipolysis in adipocytes' and 'PPAR signaling pathway', with the core target of AKT1 or PPARG ( Figure 4A). In addition, CluePedia illustrated the layout of the disposition of major pathways and their component genes according to their cellular layer of localization ( Figure 4B).

BP and KEGG Enrichment Analysis
For a detailed analysis of the 41 targets, two clusters divided using Cytoscape MCODE app were analyzed to obtain information on the pathways of the candida targets through the KEGG pathway analysis in the DAVID database, and each cluster wa represented using the bubble plot of the GGPlot2 package ( Figure 5). A and B are th results of cluster 1, confirming that they had significant results for cholestero triglyceride, the PPAR signaling pathway, and cholesterol metabolism. C and D are th results of cluster 2. Significant results were confirmed for lipid response, non-alcohol fatty liver, and atherosclerotic disease.

BP and KEGG Enrichment Analysis
For a detailed analysis of the 41 targets, two clusters divided using Cytoscape's MCODE app were analyzed to obtain information on the pathways of the candidate targets through the KEGG pathway analysis in the DAVID database, and each cluster was represented using the bubble plot of the GGPlot2 package ( Figure 5). A and B are the results of cluster 1, confirming that they had significant results for cholesterol, triglyceride, the PPAR signaling pathway, and cholesterol metabolism. C and D are the results of cluster 2. Significant results were confirmed for lipid response, non-alcoholic fatty liver, and atherosclerotic disease.

Visualization of the Target-Chemical Interaction Using STITCH
As the enrichment analysis results from the functional cluster indicated the significant potential of targets on lipid disorders, this study investigated the chemical-target network in the STITCH database ( Figure 6). As a result, the targets from cluster 1 harbor strong interactions with various currently used statins.

AT, PC, ZO, and Mixed Extract (MIX) Improved the Energy Metabolism-Related Proteins in the Hepatic Steatosis Model
The HepG2 cells did not show any significant decrease in viability at <50 µg/mL (93.96%, 96.37%, 95.84%, and 95.42% at 50 µg/mL of AT, PC, ZO, and MIX, respectively) ( Figure 7A). The influence on the protein expression of ACC, AMPK, AKT, and CPT-1 was investigated in the steatosis model. At 50 µg/mL, AT and PC stimulated the phosphorylation of ACC, AMPK, and AKT and increased CPT-1 expression related to fatty acid oxidation. In addition, ZO promoted the phosphorylation of AMPK and AKT, and the expression of CPT-1. In particular, MIX strongly upregulated AKT, AMPK, and CPT-1, comparable to the effects of AT, PC, and ZO at 50 µg/mL ( Figure 7B). MCODE app were analyzed to obtain information on the pathways of the candidate targets through the KEGG pathway analysis in the DAVID database, and each cluster was represented using the bubble plot of the GGPlot2 package ( Figure 5). A and B are the results of cluster 1, confirming that they had significant results for cholesterol, triglyceride, the PPAR signaling pathway, and cholesterol metabolism. C and D are the results of cluster 2. Significant results were confirmed for lipid response, non-alcoholic fatty liver, and atherosclerotic disease.

Visualization of the Target-Chemical Interaction Using STITCH
As the enrichment analysis results from the functional cluster indicated the significant potential of targets on lipid disorders, this study investigated the chemicaltarget network in the STITCH database ( Figure 6). As a result, the targets from cluster 1 harbor strong interactions with various currently used statins. Figure 6. Visualization of the target-chemical interaction of targets from cluster 1 using the STITCH database.

AT, PC, ZO, and Mixed Extract (MIX) Improved the Energy Metabolism-Related Proteins in the Hepatic Steatosis Model.
The HepG2 cells did not show any significant decrease in viability at <50 μg/mL (93.96%, 96.37%, 95.84%, and 95.42% at 50 μg/mL of AT, PC, ZO, and MIX, respectively) ( Figure 7A). The influence on the protein expression of ACC, AMPK, AKT, and CPT-1 was investigated in the steatosis model. At 50 μg/mL, AT and PC stimulated the phosphorylation of ACC, AMPK, and AKT and increased CPT-1 expression related to

AT, PC, and ZO Regulated the Expression of Genes Related to Lipogenesis and Reduced FFA Induced Intracellular Lipid Accumulation
The influence of herbal extracts on the gene expressions of CEBPA, PPARG, an HMGCR was investigated in FFA-induced HepG2 cells incubated with variou concentrations of AT, PC, ZO, and MIX for 6 h or 48 h. HepG2 cells stimulated with 1 mM FFAs showed increased CEBPA, PPARG, and HMGCR mRNA expressions. AT and ZO prevented these upregulations. The mixture of AT, PC, and ZO regulated these, excludin the mRNA expression of HMGCR ( Figure 8A). Furthermore, Oil Red O stainin demonstrated the superior efficacy of MIX in reducing lipid accumulation both in optica density and in the microscopic analysis ( Figure 8B).

Discussion
The holistic effects of herbal medicines were difficult to analyze because of their complex multi-components in extracts [54]. As therapeutic efficacy is based on the combined effects of mixed compounds, it was necessary to understand the complex interaction between multi-compounds from the herbal medicines and multi-targets of its compounds [55]. On the other hand, drug discovery is challenging and time-consuming when using conventional experiment-based screening methodologies [56].
In recent studies of oriental medicine, however, the network pharmacological approach allowed an understanding of the complex mechanisms of activity exerted by multiple compounds and an ability to predict the pharmacological activity [57,58]. This study deduced the targets, estimated to have a significant impact on hyperlipidemia based on the in silico study using network pharmacology analysis, and validated its real impact by performing an in vitro study. Several studies investigated the impact of natural products on hyperlipidemia using network pharmacology [15,59,60], but studies investigating herbal prescriptions (or their combinations) are relatively rare. In the present study, herbal combination exerted stronger activity than single herbs on hyperlipidemia and their related metabolic disorders, at least in certain aspects. Therefore, this result will arouse the interest of fellow researchers to conduct investigations on other herbal prescriptions on hyperlipidemia.
Three major herbs with potential use in hyperlipidemia and its backgrounding metabolic disorders were deduced by analyzing the composition of the prescription based on the empirical knowledge databases of oriental medicine. Network pharmacology

Discussion
The holistic effects of herbal medicines were difficult to analyze because of their complex multi-components in extracts [54]. As therapeutic efficacy is based on the combined effects of mixed compounds, it was necessary to understand the complex interaction between multi-compounds from the herbal medicines and multi-targets of its compounds [55]. On the other hand, drug discovery is challenging and time-consuming when using conventional experiment-based screening methodologies [56].
In recent studies of oriental medicine, however, the network pharmacological approach allowed an understanding of the complex mechanisms of activity exerted by multiple compounds and an ability to predict the pharmacological activity [57,58]. This study deduced the targets, estimated to have a significant impact on hyperlipidemia based on the in silico study using network pharmacology analysis, and validated its real impact by performing an in vitro study. Several studies investigated the impact of natural products on hyperlipidemia using network pharmacology [15,59,60], but studies investigating herbal prescriptions (or their combinations) are relatively rare. In the present study, herbal combination exerted stronger activity than single herbs on hyperlipidemia and their related metabolic disorders, at least in certain aspects. Therefore, this result will arouse the interest of fellow researchers to conduct investigations on other herbal prescriptions on hyperlipidemia.
Three major herbs with potential use in hyperlipidemia and its backgrounding metabolic disorders were deduced by analyzing the composition of the prescription based on the empirical knowledge databases of oriental medicine. Network pharmacology analysis revealed the key compounds from the herb combination that can modulate multiple targets critical to certain phenotype differentiation in oriental medicine for hyperlipidemia. In detail, a seemingly complex target network (PPI network) was divided into two different functional clusters of targets ( Figure. 2). Many common genes, which are closely related to the pathologic mechanisms of hyperlipidemia were shared by two KEGG pathways 'regulation of lipolysis in adipocytes' and 'PPAR signaling pathway', as investigated by ClueGO.
The lipolysis process can be defined as the hydrolysis of triacylglycerol into fatty acids and glycerol to be used as an energy source [61]. This result might explain the observation of decreased neural lipid accumulation by three herbal extracts in the hepatic steatosis model, as demonstrated by the Oil Red O staining assay ( Figure 7B).
The PPAR signaling pathway plays a key role in treating dyslipidemia by modulating the lipoprotein metabolism, which is targeted by fibrates [62,63]. Furthermore, the KEGG enrichment analysis of the functional clusters demonstrated its relationship with cholesterol metabolism and non-alcoholic fatty liver disease ( Figure 4B,D). The enriched BP term results of these clusters are limited to cholesterol-related terms, as well as triglycerides, sterol, and lipid homeostasis ( Figure 4A). This appears to be relative merit of the herbal combination to current statin treatment, of which its primary drug mechanism is to inhibit the ratelimiting enzyme for cholesterol synthesis [64]. In addition, five significantly enriched KEGG pathways (p < 0.05), which are strongly involved in lipid metabolism within the rank of the top 20, were found ( Table 2).
The combinational effect was evaluated by comparing the efficacy of single extracts and mixed extracts. As shown in the in vitro data, a mixed extract of three herbs and single-herb extracts against FFA-induced hepatic steatosis had a notable effect on the lipid biosynthesis genes (Figure 7). Moreover, as a complex mixture of bioactive compounds, it has several additional benefits on energy expenditure (by phosphorylating AMPK) or lipid catabolism (by increasing CPT1a expression) in the aspects of treating dyslipidemia-related complications ( Figure 6). On these markers, even more favorably, the mixed extract exerted an enhanced efficacy compared to single-herb extracts. This is more advantageous when patients with hyperlipidemia usually have comorbidities of other metabolic disorders related to obesity [65,66]. In addition, the STITCH result suggested that the targets of the herbal combination have a close interaction with several statins that might imply a similar drug mechanism. Therefore, the herbal mixture might provide an option to treat hyperlipidemia with new molecular targets and mechanisms that can ameliorate metabolic disorders and ultimately alternate statins.
However, this study presented limited evidence of the possibility of the herbal combination on hyperlipidemia and other metabolic disorders. The 41 key target genes do not include the core targets related to the lipid metabolism, such as SREBF1 and SCD1 (Table 2). Moreover, not all the targets could be validated in the experiments deduced from network pharmacology. The herbal combination sample might show efficacy in the steatosis model but not in hyperlipidemia because it could not be determined that the reduced accumulation of lipid droplets was not caused by decreased fatty acid uptake in media.
The hypolipidemic efficacy of the herbal combination needs to be validated in in vivo studies. In particular, a comparative study to evaluate the overall efficacy of the herbal combination on metabolic disorders, not only for dyslipidemia, should be conducted in comparison with the current statin doses.

Conclusions
The pharmacological activities of three herbs, including ZO, AT, PC against dyslipidemia, were estimated via network pharmacological approaches. The herbs potentially target several key proteins and pathways critical to lipid metabolism. The efficacy and mechanism of the herbal extracts were investigated in hepatic steatosis model in vitro. The mixed herb extract showed a stronger potential against hepatic steatosis compared to single-herb extracts. As a result, we suggest that the herbal combination could be a candidate drug for hyperlipidemia which can alternate statin, with significant benefits on modulating lipid metabolism.