Polypharmacology-based Approach for Screening of Traditional Chinese Medicinal extract and Multi-dimensional Targeting Verication

Background: Natural products and their unique polypharmacology offer signicant advantages for nding novel therapeutics particularly for the treatment of complex diseases. Meanwhile, natural products in traditional Chinese medicine possess drug-like properties. In this study, we used the previously established co-infection model of Mycoplasma gallisepticum and Escherichia coli as a representative of complex diseases. The multi-omics joint analysis was used to reverse screen TCMs from the Chinese medicinal database and then targeted verication was conducted from multiple dimensions. Results: The results showed that the Chinese herbal compound screened by the target network played a good therapeutic effect in the case of co-infection. Furthermore, the four methods were performed at gene, protein, and metabolite levels, as well as molecular docking in vitro respectively, which were used to verify the multi-target therapeutic effect. Conclusions: These data suggest that we may provide a new method to validate target combinations of natural products that can be used to optimize their multiple structure-activity relationships to obtain drug-like natural product derivatives. Furthermore, the study could establish a new methodology for the research of Chinese herbal medicinal extract and provide a new multi-target treatment method for the occurrence of co-infection. Genomes; CCU, color change unit; UPLC, ultra-performance liquid chromatography; MS, mass spectrometry; MS/MS, tandem mass spectrometry DEMs, differentially expressed metabolites; DEGs, differentially expressed genes; SD, standard deviation; TCM, traditional Chinese medicine; NCMC, new Chinese medicinal compound; PPI, Protein-protein Interaction; MIC, minimum inhibitory concentration; BDKRB1, B1 bradykinin receptor; CHRM4, Cholinergic receptor muscarinic 4; EDN1, Endothelin-1; FOS, Cellular oncogene fos; HRH3, G protein coupled receptor 97; IFNG, Immune interferon receptor 1; MMP2, Matrix metalloproteinase 2; MYC, Myc proto oncogene protein; NOS2, Nitric Oxide Synthase 2; TLR4, Toll like receptor 4; VAMP2, Vesicle Associated Membrane Protein 2; VEGFA, Vascular endothelial growth factor A.


Background
Polypharmacology has been widely recognized as a new approach of modern drug discovery [1,2], which might enable natural products to surpass the use of the traditional single-target drugs in terms of e ciency [3]. Natural products with polypharmacological pro les show a promising prospect for the development of novel therapeutics for various complex diseases [4]. The methods commonly used in polypharmacology include Molecular docking [5,6], Network-based approaches [7,8], and Omics-based systems biology approaches [9][10][11] along with the development of in silico pharmacology [12]. However, unlike pharmacological drugs used in Western medicine, the active components of natural medicines are often not precisely speci ed [13], which has greatly restricted the development of traditional Chinese medicine (TCM). But more and more research methods on Chinese herbal compounds have been reported in recent years. Pan's study focused on the intervention of Huanglian Decoction in rats with type 2 diabetes mellitus by the methods of network pharmacology and metabolomics [14]. Zhou probed into the effect of Liuwei Dihuang decoction on the neuroendocrine immunomodulation network [15]. In the present study, network pharmacology and systems biology will enable us to deliberately design lead molecules with expected polypharmacology as well as offer opportunities for TCM formula repurposing.
Due to the intensi cation of commercial poultry production, explosive multiple respiratory infections have become an urgent problem [16]. This co-pathogenesis is characterized by complex interactions between co-infection pathogens and the host [17,18]. Combined with the previous researches on the co-infection of Mycoplasma gallisepticum (MG) and Escherichia coli (E.coli) in our laboratory [16,19,20], we try to screen a new Chinese medicinal compound (NCMC) with the method of multi-pharmacology and carry out targeted veri cation. There are many examples of natural products being used in drug discovery efforts that are directed at a wide range of indications beyond their traditional strengths with the development of research technologies [21][22][23]. Moreover, the complex mechanisms of co-infection were gradually unraveled [24,25]. However, whether the co-infection can be combined with the complex effects of TCM compounds has never been studied. The study could establish a new methodology for the research of Chinese herbal medicinal extract and provide a new multi-target treatment method for the occurrence of co-infection.

Materials And Methods
Bacterial infection and Experimental groupings R low strain was obtained from the Harbin Institute of Veterinary Medicine (Chinese Academy of Agricultural Sciences), which was grown in a Modi ed Hay icks medium. Escherichia coli O78 was isolated from chickens infected with colibacillosis in our laboratory and cultured in Nutrient Broth (Beijing Aoboxing BIO-TECH Co., Ltd.). The concentration of MG and E.coli were 1 × 10 9 CCU/ml and 10 9 CFU/ml respectively. The detection of the density for MG and E.coli were consistent as explained in our previous study [19,26,27].
Forty (1 day old) commercial Leghorn chickens were obtained from Chia Chau Chicken Farm (Harbin, Heilongjiang, China) and were assigned randomly to 2 groups namely the Control group and Co-infection group as described in our earlier study [19], Control group: Fed only basal diet; Co-infection group (B): 0.2 ml of MG medium (1 × 10 9 CCU/ml) was injected into the left caudal thoracic air sac on the 7th day, and 0.1 ml of E.coli bacteria (10 9 CFU/ml) was injected intraperitoneally on day 10. On the 13th day, 20 chickens from each group were euthanized using the method of cardiac blood collection. Lung samples in each group were collected for RNA-seq, while the serum samples were collected for non-targeted metabolomics.
Collection of co-infection target genes and construction of TCM-target network RNA was extracted by Trizol reagent (Invitrogen Inc., Carlsbad, CA) from lung tissue and was utilized to construct the nal library (BGISEQ-500 RNA-Seq Library) [19]. DEG-seq method was based on the Poisson distribution (Fold Change > 2 and Adjusted P-value < 0.001) [28,29]. According to the KEGG annotation results and the o cial classi cation, we separately classi ed the functional and biological pathways of the DEGs. The DEGs from different sub-categories were compared to the STRING database by DIOMAND (www.diamondsearch.org/). To obtain PPI (Protein-protein Interaction) diagram, the gene interactions were obtained by homology with known proteins. The network relations with a score ≥ 300 were screened out for mapping. After that, the genes with the highest score in PPI results were selected as target genes and used for subsequent network construction.
In the process of network construction, we rst collected all the information of the TCMSP database (https://tcmspw.com/tcmsp.php) and got the total network of "TCM-component-target", which was submitted in Supplementary Material 2. We then put the collected target genes information into the total network for the reverse screening of TCM. The TCM was sorted according to the correlation degree of nodes in the reverse screening network. Furthermore, we classi ed the TCM based on different functions (Chinese Pharmacopoeia 2015 Edition) and chose the top1 of each classi cation, including Isatdis Radix, Forsythia Fructus, Ginkgo Folium, Mori Cortex, Licorice, and Radix Salviae. The details for these single Chinese medicinal herbs and materials and their roles based on TCM theory are listed in Supplementary Material 8. Finally, the TCM-component-target network was established using Cytoscape 3.6.1 software (Bethesda, MD, USA).

Preparation of NCMC
To verify whether the selected compound has a better therapeutic effect and determine the proportion of six herbs. Hence, we had made a treatment experiment based on the uniform design with the methods of pharmacodynamics and minimum inhibitory concentration (MIC). The results showed that the ratio of Isatdis Radix, Forsythia Fructus, Ginkgo Folium, Mori Cortex, Licorice, and Radix Salviae were 14:7:11:12:5:3 showed effective treatment. The design optimization details of NCMC for the treatment of co-infection was provided in Supplementary materials 7.
Six herbs were purchased from Runhe Chinese medicine processing plant Ltd. Aqueous extract of NCMC was prepared as the following procedure. The medicinal materials were mixed in proportion and were macerated for 1 h with ten folds distilled water (v/w), and then decocted for 1 h, after which the ltrate was collected and the residue was decocted again for 1 h up to six folds (v/w) in distilled water. The ltrates were mixed and condensed and then dried by vacuum-drier at 60 ℃ [30,31]. The nal concentration of the aqueous extract is 1 mg/mL. UHPLC-QTOF-MS analysis for components quanti cation 100 µL sample was added to 400 µL of an extracted solution containing 1.25 µg/mL of internal standard which dissolved in water. After 30 s vortex, the samples were sonicated for 10 min in an ice-water bath. After the samples were incubated at -40 ℃ for 1 h, then the sample was centrifuged at 1,2000 rpm for 15 min at 4 ℃. Finally, the supernatant was put in a fresh 2 mL tube and 200 µL was transferred to a fresh glass vial for LC-MS/MS analysis. LC separation was performed on the Nexera UHPLC LC-30A system (SHIMADZU, Japan) with a Waters BEH C18 column (1.7µm*2.1*100 mm, Waters, USA). Mobile phases, water with 0.1% formic acid (A) and acetonitrile (B), were applied with gradient elution and the ow rate was kept at 0.4 mL/min. AB 5600 Triple TOF system (SCIEX, USA) was used to collect primary and secondary MS data based on IDA function under the control software (Analyst TF 1.7 (AB Sciex). Instrument dependent parameters: curtain gas = 35 psi, IonSpray voltage = + 5500 (POS)/-4000 (NEG) V, nebulizer gas = 55 psi, heater gas = 55 psi, source temperature = 550 ℃. The original mass spectrometry data was imported using Progenesis QI software. The corresponding TCM metabolic database in the compound was established, and the peaks containing MS/MS data were identi ed by the self-built secondary mass spectrometry database (Biotree Biomedical Technology Co., LTD, China).

NCMC treatment and groupings
A total of 120 White Leghorn chickens was purchased from Chia Chau Chicken Farm (Harbin, Heilongjiang, China), which were divided randomly into 4 groups as follows. Control group (A): Chickens in the CG were fed only with basal diet and inoculated with the culture medium 0.2 mL on day 7 in air sac; Co-infection group (B): The co-infection model was constructed as previously; Co-infection + NCMC administration group (C): The same infection model as mentioned, then treated with the aqueous extract of NCMC, and given orally by gavage. The treatment started on day 13 and continued for 5 days, once a day at a dose of 450 mg/kg; NCMC control group (D): The same dose of NCMC (450 mg/kg) was given orally to chickens by gavage started at day 13 and continued for 5 days. On the 18th day, 20 chickens from each group were euthanized using the method of cardiac blood collection. Lung, tracheal, and serum samples in each group were collected for further analysis.
Microscopic examination of the trachea First, the tracheal tissues were xed in 10% formalin, dehydrated, and immersed in the transparent samples of wax, cut into slices (4 µm), stained with hematoxylin and eosin (H & E). Secondly, the tracheal tissues were trimmed into small pieces of 1 mm 3 and xed overnight in 2.5% glutaraldehyde. They were washed with PBS twice and post-xed in 1% osmium tetroxide at 4 ℃ for 1 h. Next, the tissues were dehydrated by ethanol series and 100% acetone, embedded in epoxy resins. The ultrathin sections were stained with uranyl acetate and lead citrate and then observed under a GEM-1200ES transmission electron microscope (JEOL Ltd., Tokyo, Japan). Lastly, 1 mm 3 piece of tracheal tissues was also observed by scanning electron microscopy (SEM, SU8010, HITACHI Ltd., Japan). Joint pathway analysis 16 serum samples of Control and Co-infection groups were examined by the LC-MS system as explained in our previous study [16]. Differentially expressed metabolites (DEMs) were screened in combination with univariate analysis of fold change and q-value. Screening conditions: 1) VIP ≥ 1; 2) fold change ≥ 1.2 or ≤ 0.8333; 3) q-value < 0.05. The DEMs and the previous target gene symbols were imported into the MetaboAnalyst (MetaboAnalyst v4.0), which could simultaneously analyze genes and metabolites of interest within the context of metabolic pathways [32,33].

Quantitative RT-PCR and western blotting of targets genes
The lung tissue samples were homogenized for 2 min at a low frequency of 65 Hz using an automatic tissue homogenizer machine (Shanghai Jingxin Industrial Development Co., Ltd.). Total RNA was extracted using Trizol reagent (Invitrogen Inc., Carlsbad, CA, United States) and the reverse transcription of cDNA was performed according to the manufacturer's instructions (Takara Biomedical Technology (Beijing) Co., Ltd.). The primer sequences are shown in Table 1. Quantitative RT-PCR was performed to analyze gene expression using a LightCycler96 (Roche, Basel, Switzerland). Each sample was analyzed in triplicate. The fold change in gene expression was calculated using the △△cycle time (Ct) method after the expression level was normalized with the GAPDH gene taken as an internal standard. Western blotting was used to measure the related target proteins. Total proteins of lung tissues were extracted by the whole-cell lysis assay and the supernatant protein content was determined using the BCA protein assay kit (Wanlei, Liaoning, China). The membranes were incubated overnight on a shaker at 4 ℃ with primary antibodies against β-actin (1:5000 dilution), MMP2 (1:500 dilution), TLR4, c-FOS, BDKRB1, VEGFA, and EDN-1 (1:1000 dilution). All the primary antibodies purchased from Bioss Bioscience Inc. (Beijing, China), and species homology included Gallus gallus. Secondary anti-mouse and anti-rabbit IgG peroxidase were used for 1 h, and then bound immune-complexes were detected using enhanced chemiluminescence (ECL) detection. The protein bands were analyzed by densitometry using Image J (V 1.42, National Institutes of Health, USA).

Detection of major metabolites by LC-MS
Three key metabolites (Dopamine, γ-Aminobutyric acid, and Leukotriene C4) were obtained from the results of the joint analysis, and the LC-MS method was used for the detection and quanti cation of targeted metabolites in the anion mode. We rst prepared the serum samples and standards for pretreatment by organic precipitation. The information of the three metabolite standards is as follows, Dopamine (BD161397, Bide Pharmatech Ltd., Shanghai, China), γ-Aminobutyric (P10002, APEBIO, USA), Leukotriene C4 (GC18663, GLPBIO, USA). The mobile phases of Dopamine and γ-Aminobutyric acid were A: 20 mM ammonium formate solution (0.1% formic acid), B: methanol. While, the mobile phases of Leukotriene C4 were: A (0.1% formic water), B (0.1% acetonitrile formate). Then, the Agilent 6490 QQQ mass spectrometer (Agilent Technologies, England) was conducted to obtain the primary and secondary mass spectrometer data based on MRM mode and adopted positive mode. The parameters of the ESI ion source were set as follows: As the 3D structure of these ve proteins (Gallus gallus) has not been elucidated yet, the method of comparative modeling was used for their 3D structure [35,36] to construct the model using the alignment mode. We used PyMol [37] to erase the heteroatoms, water molecules, and inhibitor present in the structure and saved as a PDB le. The 3D structures of the ligand molecule were stored as a Mol2 or SDF le with the help of the TCMSP databases [38] and PubChem (https://pubchem.ncbi.nlm.nih.gov/). The non-bonding interaction of ligand-protease was calculated using Autodock Vina software package [39] for docking analysis. After docking, the interaction of the ve proteins and Baicalin with the lowest a nity score for the receptors were selected for further analysis as mentioned in our previous study [40].

Statistical analysis
Data are presented as mean results ± standard deviation (SD). All the experiments were performed in triplicates (n = 3) unless otherwise mentioned. The signi cance was determined using one-way ANOVA followed by Dunnett's T3 test. The data were analyzed by using the GraphPad Prism (version 5.01). Values with p < 0.05 were considered statistically signi cant. Heatmaps were made by Heatmap Illustrator software (1.03.7). The KEGG classi cation and the PPI maps were made by the BGI data mining online website (http://report.bgi.com). NCMC component classi cation diagram was made by RAW graphs (https://rawgraphs.io/).

Results
Transcriptome sequencing of co-infection group and targets screening In this study, multi-omics analyses were performed on the co-infection model as explained previously [19]. According to the results of RNA-seq, 3,115 DEGs were found between the co-infection and the control groups, including 1,456 genes upregulated and 1,659 genes downregulated. These DEGs were widely distributed in 44 sub-categories in six major categories in the KEGG pathway database, as shown in Fig. 1A. The PPI results in each KEGG classi cation were submitted in Supplementary Material 1, and the target genes obtained from screening were summarized in Fig. 1A.

Reverse screening of NCMC in TCMSP database
We rst collected the entire data of the TCMSP Chinese Medicine Database, which were submitted in Supplementary Material 2. 57 target genes were substituted into the Drug bank database and 25 effective targets were obtained. These 25 targets were brought into the TCMSP total network for the reverse screening of Chinese medicine, and the results showed that 15 targets were matched with the TCMSP database (Reverse network as shown in Supplementary Material 3). Afterward, we sorted the correlation degree of Chinese medicine nodes and selected six high-ranked Chinese medicines according to the classi cation of Chinese medicines. The six herbs were Isatdis Radix, Forsythia Fructus, Ginkgo Folium, Mori Cortex, Licorice, and Radix Salviae, and the nal "target-component-TCM" network was shown in Fig. 1B.

Component compounds of NCMC aqueous extraction
To clarify the effective ingredients of NCMC, a non-targeted metabolomics method was used to study the aqueous extraction. The results showed that 260 compounds (including positive and negative modes) were detected on the receiver side using UPLC-QTOF-MS/MS. Among them, 85 compounds belong to avonoids, 64 compounds belong to organic acids, and 17 compounds belong to phenylpropanoids, and so on, as shown in Fig. 2. The complete composition data of NCMC is provided in Supplementary Material 4. Furthermore, the role of the effective ingredients of NCMC was still not clear. Hence, we have screened effective compounds with higher content from the total ion current in both the positive and negative ion modes as shown in Fig. 3. The 2D structure of the ingredients was obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) and saved as SDF les for subsequent experiments.
Pharmacodynamics' evaluation of NCMC Pathological and ultrastructural changes (Fig. 4) were performed to better understand the effects of NCMC treatment on the chicken trachea. In the co-infection group (Fig. 4A), the cilia are exfoliated and there is a proliferation of goblet cells in the upper mucosa. While the cilia are well arranged and there is still a small amount of in ammatory cells in ltrations after NCMC treatment (Fig. 4B). Scanning electron microscopy (SEM) showed that the cilia were signi cantly more abundant and intact after treatment of co-infection ( Fig. 4C and 4D). We can see that cilia rupture, inverted, cytoplasm swelling, matrix electron density decreased, and the formation of umbrella-like processes to the extracellular in the co-infection group by TEM (Fig. 4E). Although some cilia were shed, the symptoms of the treatment group were signi cantly improved compared with the co-infection group (Fig. 4F).

Multi-omics joint analysis of co-infection
To explore the metabolic alterations associated with co-infection, we used a UPLC-QTOF-MS/MS-based metabolomics method to examine metabolite alterations in the serum. The non-targeted metabolomics results of the co-infection group showed that there were 245 (58 upregulated and 187 downregulated) and 882 (234 upregulated and 648 downregulated) metabolites in positive and negative modes respectively [16]. Meanwhile, the previous 57 key target genes and differential metabolites were introduced into MetaboAnalyst (v4.0) for further multi-omics joint analysis. The gene-metabolite network is shown in Fig. 5A. Among them, we selected the metabolites and genes with the largest node association degree for subsequent multi-dimensional targeted veri cation.

Effects of NCMC on the expression of genes and proteins in the lung
According to the gene-metabolite network, 12 genes were screened to investigate the underlying mechanism of the positive effect of NCMC on the target organ, which was used to draw a heatmap for the relationships between the three groups, as shown in Fig. 5B. The heatmap shows that there are different degrees of the eruption of chemokines and mucins secreted by the Co-infection group, except CHRM4, MMP2, and VEGFA. However, half of the genes in the Co-infection + NCMC group showed a signi cant decrease.
Six key proteins were selected in the network for WB experiments to conduct in-depth research, namely BDKRB1, EDN1, MMP2, TLR4, c-FOS, and VEGFA. From the results of these six proteins ( Fig. 5C and 5D), we found that most of the proteins were signi cantly (0.01 < p < 0.05) unregulated in the co-infection group compared to the control groups, except c-FOS. In particular, EDN1, MMP2, and VEGFA showed extremely signi cant (p < 0.01) expressions. After NCMC treatment, the expressions of these target proteins was signi cantly downregulated in contrast to the Co-infection group. From our results, NCMC signi cantly modulated the expression of some genes and/or the corresponding proteins, such as MMP2 and TLR4 (p < 0.01).

Expression of NCMC on the key metabolites in the serum
We have previously demonstrated that LTC4 in serum could be used as the biomarker for detecting poultry respiratory diseases [16]. Combined with the results of transcriptome-metabonomic network analysis, we, therefore, chose these three metabolites (Dopamine, γ-Aminobutyric acid, and Leukotriene C4) for the detection of the metabolic level. As shown in Fig. 5E and 5F, the expression of three metabolites was up-regulated and extremely signi cant (p < 0.01) between the Control group and the Coinfection group. However, Dopamine and γ-Aminobutyric acid in the NCMC treatment group showed signi cant reduction (p < 0.05) compared with the Co-infection group. Strikingly, the expression of LTC4 showed no signi cance even after treatment. The original gures (including western blot and UPLC/MS) were provided in Supplementary Material 5.

Molecular docking of effective compounds and target proteins
Based on the results of non-targeted metabolomics of NCMC, a total of 21 effective compounds (including NEG and POS modes) were sorted out. We then performed the molecular docking of the 21 compounds and 12 key proteins. The compounds with the highest docking score absolute value of all 12 key proteins were selected as candidate inhibitors (Docking original gures and docking score details were uploaded as Supplementary Material 6). The optimum ligand-receptor complexes were generated by PyMol and the interaction analysis was analyzed by Protein-Ligand Interaction Pro ler [41], as shown in Fig. 6. The most adaptable results of molecular docking showed that BDKRB1 (-6.7 kcal/mol), CHRM4 (

Discussion
In this study, we conducted multi-target screening for co-infection of MG and E.coli based on omics research and investigated the polypharmacology mechanism of NCMC with multiple components. The whole research were based on the establishment of a co-infection model and the subsequent target screening. According to the previous transcriptome studies on co-infection [19], we obtained a total of 3,115 DEGs and the corresponding KEGG classi cation. As the previous study showed that oxidative stress transcriptomics and drug discovery approaches could identify and target neurotoxic innate immune populations and lead to the development of selective neuroprotective strategies [42]. However, we hope to screen the representative DEGs from the overall level of in vivo studies. Hence, the DEGs from 33 sub-categories of KEGG have carried out PPI analysis respectively as explained in previous studies [43][44][45]. The screened target genes cannot be applied directly but are matched to the corresponding target according to the Information in the DRUG BANK database.
Hopkins proposed the concept of network pharmacology in 2007, which aims to design a new generation of drugs by incorporating biological networks rather than single target [46]. Network pharmacology-based strategy to predict therapeutic targets of tanshinone I and cryptotanshinone against in ammation, and further to investigate the pharmacological molecular mechanism in vitro [47]. For complex diseases, the method of applying network pharmacology was also used to reveal the mechanism of the action of TCM [48,49]. We collected the whole data of TCMSP for the construction of the total network "TCM-component-target", and then the targets were substituted into the total network for the reverse screening of TCMs. This method of reverse operation is similar to existing relevant studies [50,51]. However, this is the rst attempt to apply the reverse screening method directly to the TCM network.
After obtaining the order of Chinese herbs, we then selected different categories based on the compatibility theory of Chinese medicines. According to the degree of node correlation, the order was as follows: Isatdis Radix (4th, yang tonic) [52], Forsythia Fructus (11th, Antipyretic and antidote) [53], Ginkgo Folium (5th, Antitussive and antiasthmatic drugs) [54], Mori Cortex (3rd, Antitussive and antiasthmatic drugs) [55], Licorice (1st, reinforcing drugs) [56] and Radix Salviae (2nd, Drugs with the e cacy of modifying rheological properties of blood) [57]. Aqueous extract of Chinese medicinal compound is the most common method to study compound components [58,59]. Hence, we sought points that were uniformly scattered on the domain [60,61], furthermore, we did a multi-nonlinear regression analysis to search for the optimal combination in preliminary experiments. Finally, we obtained the optimal combination of NCMC aqueous extract for subsequent targeted therapy experiments. In the treatment experiment of NCMC aqueous extract for co-infection, we used multiple pathological methods to observe whether a certain curative effect was achieved. Pathological and ultrastructural results showed that the symptoms of co-infection were signi cantly relieved after treatment, which indicated that NCMC played a signi cant therapeutic role.
The multi-dimensional validations of targeted therapy were performed at gene, protein, and metabolite levels, as well as computer simulations in vitro, respectively. Afterward the joint analysis of transcriptomics and metabolomics, we obtained the network map of key genes and metabolites. Recent studies show that combined metabolomics and transcriptomics analyses provide sensitive approaches to link infection to biological responses [62,63]. In our study, the rst 12 target genes with the highest correlation degree were selected for RT-PCR quantitative detection. The results showed that the expressions of these related genes were reduced after treatment. In particular, various studies have shown that these genes controlled a broad and exible network in the transcriptome [64,65]. However, the expression of CHRM4, MMP2, and VEGFA was shown opposite results. We performed western blot to verify protein level expression, which showed similar trends of the six proteins (MMP2, TLR4, c-FOS, BDKRB1, VEGFA, and EDN-1). The quantitative results of targets indicated that the NCMC aqueous extraction can play a multi-target therapeutic role after the reverse screening by network pharmacology.
In the circulation, as the basis of organism phenotypes, metabolites can help us to understand life phenomena more intuitively and effectively and reveal the essence of life [66,67]. Our previous metabolic results demonstrated that Arachidonic Acid metabolism is activated in co-infection, and it is recommended that LTC4 in serum acts as a biomarker for detecting poultry respiratory diseases [16]. Based on the results of a joint analysis of transcriptomics and metabolomics, dopamine, γ-Aminobutyric acid, and leukotriene C4 were selected for further study. It had been shown that bacterial infection can cause a dopamine burst in the brain and Dopamine activates NF-κB and primes the NLRP3 in ammasome in primary human macrophages [68,69]. The study also unveiled that γ-Aminobutyric acid, the principal inhibitory neurotransmitter in the brain, has activation functions in the immune system [70]. These ndings are similar to the results of co-infection. Furthermore, the expression of these three metabolites decreased signi cantly after NCMC treatment.
Finally, to further verify which active components of NCMC aqueous extraction have a multi-target effect, molecular docking was used for in vitro analysis. A nity was the score for the molecular docking, and when the score was lower, the binding a nity was stronger. An a nity < − 7 indicated a stronger binding activity [39]. Our docking results demonstrated the multi-target action of the main active components of the NCMC. In particular, the targeted therapy of Salvianolic acid A, Liquiritigenin, and Isoliquiritigenin could be further investigated. There are even more naturally derived compounds that are either in clinical trials or have shown potential in pre-clinical studies which merit further investigation [71]. In the present study, we used four methods to verify the multi-target therapeutic effect of NCMC extraction. It can be found that NCMC plays a certain role in the expression of key target proteins, and the corresponding key metabolites have also been con rmed.
However, several potential limitations in our experimental and polypharmacology-based approaches should be acknowledged. First, the incomplete known components of natural products curated from publicly available databases limit our approach to predict NCMC extraction for those without known information. Furthermore, the MetaboAnalyst (v 4.0) was used to simultaneously analyze genes and metabolites of interest within the context of metabolic pathways. But only data from humans, mice, and rats are supported currently [32]. Integrating the target network may help to target the growing potential of co-infection genes via indirectly acting on their neighbor proteins in the human interactome. Lastly, the development of TCM should focus on interdisciplinary, combining research from emerging elds after clarifying the material basis. In terms of pharmacodynamics, it is necessary to clarify how NCMC extraction works, such as analyzing the dialectical relationship between "Chinese herbal medicine, probiotics, and intestinal ora".

Conclusion
In summary, the target network for the co-infection was dug through the joint analysis of transcriptomics and metabolomics, and the mapping between the target network and the total TCM network was used to reverse screen TCMs. We started from the network and carried out targeted veri cation from multiple dimensions to preliminarily reveal the polypharmacological effects of NCMC. These data suggest that we may provide a new method to validate target combinations of natural products, which perhaps to optimize their multiple structure-activity relationships to obtain drug-like natural product derivatives. Furthermore, reverse screening TCMs, which can be used to determine the functional role of TCMs in a wide range of co-infection. Thus, transcriptomics and metabolomics approaches could identify and target co-infection and lead to the development of selective polypharmacology strategies.

Consent for publication
Not applicable.

Availability of data and materials
The data analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.