Abstract Traditional Chinese medicine (TCM) is one of the oldest East Asian medical systems. The present study adopted a systems biology-based approach to provide new insights relating to the active constituents and molecular mechanisms underlying the effects of dragon's blood (DB) tablets for the treatment of colitis. This study integrated chemical analysis, prediction of absorption, distribution, metabolism, and excretion (ADME), and network pharmacology. Firstly, a rapid, reliable, and accurate ultra-performance liquid chromatography-electrospray ionization-tandem mass spectrometry method was employed to identify 48 components of DB tablets. In silico prediction of the passive absorption of these compounds, based on Caco-2 cell permeability, and their P450 metabolism enabled the identification of 22 potentially absorbed components and 8 metabolites. Finally, networks were constructed to analyze interactions between these DB components/metabolites absorbed and their putative targets, and between the putative DB targets and known therapeutic targets for colitis. This study provided a great opportunity to deepen the understanding of the complex pharmacological mechanisms underlying the effects of DB in colitis treatment. Introduction Traditional Chinese medicine (TCM) is one of the oldest medical systems of health care, and it has been used in East Asian countries such as China, Japan, and Korea for thousands of years [49][1], [50][2]. However, acceptance of traditional Chinese medicines within Western biomedical practice has been restricted by a lack of knowledge of the active compounds involved, and their therapeutic mechanisms of action [51][3]. TCM is characterized by the usage of multi-component, multi-target agents that collectively modulate molecular networks. Recently, network pharmacology has provided a means to improve understanding of the molecular mechanisms underlying the therapeutic effects of traditional Chinese medicines [52][3]–[53][5]. TCM chemical databases such as TCM Database@Taiwan [54][6] and HerbBioMap database [55][7] and some literature generally provide the main sources of information on the chemical profiles of traditional Chinese medicines. However, these are not always accurate because the chemical profiles of traditional Chinese medicines can vary significantly, depending on the geographical origin of the materials used, the harvest time, pretreatments, and manufacturing processes employed. In addition, most TCM formulations are taken orally, estimation of intestinal absorption and cytochrome P450 metabolism provide more in-depth insights into their therapeutic mechanisms [56][8], [57][9]. However, experimental determination using these systems can be costly and time-consuming. To obtain a rapid estimation of human absorption and metabolism, high throughput screening methods using Caco-2 cell monolayers and P450 enzymes provide the most advanced in vitro approaches to assessing new chemical entities [58][10]–[59][12]. Thus, the development of rapid high-throughput approaches to studying TCM compositions and mechanisms, combining chemical analysis, prediction of absorption, distribution, metabolism, and excretion (ADME), and network pharmacology is required. According to the Pharmacopoeia of the People's Republic of China, commercially available longxuejie or dragon's blood (DB) includes plant resins from four species; Dracaena spp., Daemonorops spp., Croton spp., and Pterocarpus spp., and has long been used as an ethnomedicine in China to invigorate blood circulation in the treatment of traumatic injuries, blood stasis, and pain [60][13], [61][14]. DB has also been used to treat chronic colitis in recent decades and many clinical studies have reported good therapeutic effects [62][15]–[63][17]. Chemically, flavonoids, phenols, steroides, and terpenoids have been identified as the main constituents of DB [64][18]–[65][22]. Some analytical methods have been developed to identify and characterize DB composition and assess its quality [66][23]–[67][25]. Pharmacologic studies of these components have identified a wide variety of actions, including anti-inflammatory, antiulcer, antimicrobial, hemostatic, and analgesic activities, which could contribute to their efficacy in the treatment of chronic colitis [68][26]–[69][28]. However, the precise active constituents of DB that have benefits in chronic colitis, and their molecular mechanisms of action, are still unclear. In particular, the lack of elucidation of the ingredient-target network has hindered understanding of the molecular mechanisms of DB in chronic colitis treatment. The present study employed high-throughput analysis, in silico ADME models, and a network pharmacology technique to investigate the active constituents of DB and their actions on molecular networks in chronic colitis, as shown in [70]Figure 1 . Figure 1. A schematic diagram of this systems biology-based investigation into the molecular mechanisms of DB tablet for colitis by integrating chemical analysis, ADME prediction and network pharmacology. [71]Figure 1 [72]Open in a new tab Materials and Methods 1. Reagents and chemicals High performance liquid chromatography (HPLC)-grade acetonitrile, formic acid, and methanol were obtained from Merck (Darmstadt, Germany). Water was purified using a Milli-Q system (Millipore, Billerica, MA, USA). 7,4′-Dihydroxyflavan, loureirin A, loureirin B, dracaenin A and pterostilbene standards were purchased from the National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China). The purities of all standards were no less than 98% and suitable for liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. DB enteric-coated tablets were supplied from Yunnan Datang Hanfang Pharmaceutical Co. Ltd. (Yunnan, China). 2. Preparation of samples and standard solution Six different batches of DB enteric-coated tablets were pulverized with a 60 mesh, respectively. Each sample of this powder (0.1 g) was weighed precisely and ultrasonically extracted in 10 ml methanol for 20 min at room temperature. The solution was centrifuged at 12000 rpm for 10 min, and then filtered through 0.22 µm nylon membrane filters. The filtrate was analyzed directly by UPLC-ESI-MS/MS, as described in section 2.3. At the same time, a stock solution containing five standards (7,4′-dihydroxyflavan, loureirin A, loureirin B, pterostilbene, and dracaenin A) was prepared in methanol. All solutions were stored at 4°C prior to analysis. 3. Instrument and UPLC-ESI-MS/MS conditions UPLC was performed on a Dionex UltiMate 3000 system (Dionex Corporation, Sunnyvale, USA) equipped with a quaternary pump, an online vacuum degasser, an autosampler, and an automatic thermostatic column oven. A Thermo Hyper Gold C18 column was used (100×2.1 mm, 1.9 µm) at 30°C with a flow rate of 0.4 ml/min and an injection volume of 5 µl. The mobile phase was a mixture of 0.1% formic acid in water (A) and acetonitrile (B). The mobile phase gradient program was 5% B, 0–3 min; 5–95% B, 3–20 min; 95% B, 20–25 min. High-resolution accurate-mass full scan LC-MS and LC-MS/MS analyses were performed using Thermo Q-Exactive (Thermo Fisher Scientific, Bremen, Germany). Full scans were acquired in the mass analyzer at 100–1500 m/z with a resolution of 70000 in both positive and negative ion modes, and MS/MS scans were obtained with a resolution of 17500 using a normalized collision energy of 30% for high-energy collisional dissociation fragmentation. Based on the best response for most compounds, the final parameters were set as follows: spray voltage, 3.2 kV in the positive mode and 3.0 kV in the negative mode; capillary temperature, 350°C; sheath gas pressure, 35 arbitrary units; auxiliary gas pressure, 10 arbitrary units; and heater temperature, 300°C. Xcalibur 2.2 software (Thermo Fisher Scientific) was used for data acquisition. 4. Prediction of absorbed constituents and their metabolites using in silico ADME models Structural information (*.mol or *.sdf files) relating to the DB components(identified by UPLC-ESI-MS/MS) were downloaded from ChemSpider ([73]http://www.chemspider.com/). ADME evaluation of these constituents was carried out using ACD/Percepta software 5.07 (ACD/Labs, Toronto, Canada), including the passive intestinal permeability of Caco-2 module and the P450 regioselectivity module to predict their oral bioavailability. The apparent permeability coefficient (Papp) of one constituent was greater than 9.0×10^6 cm/s, which indicated good absorption characteristics. The score and the reliability of metabolic reaction were set at greater than 0.6 and 0.5, respectively, in order to increase prediction credibility. 5. Known therapeutic targets for colitis treatment Known therapeutic targets were obtained from the DrugBank database [74][29] ([75]http://www.drugbank.ca/, version: 3.0). We only included drug-target interactions where the drugs were approved by the Food and Drug Administration, USA (FDA) for the treatment of colitis, with human gene/protein targets. To facilitate data analysis, all protein identification codes were converted to a common UniProtKB-Swiss-Prot code, and detailed target information is provided in the supplementary [76]Table S1. 6. PPI data collection PPI data were imported from eight existing PPI databases, including the Human Annotated and Predicted Protein Interaction Database (HAPPI) [77][30], Reactome [78][31], Online Predicted Human Interaction Database (OPHID) [79][32], InAct [80][33], Human Protein Reference Database (HPRD) [81][34], Molecular Interaction Database (MINT) [82][35], Database of Interacting Proteins (DIP) [83][36], and PDZBase [84][37]. Detailed information on these PPI databases is provided in the supplementary [85]Table S2. 7. Pharmacological mechanism analysis 7.1. Prediction of putative DB targets As described in our previous study [86][38], we used the Drug Similarity Search tool in the Therapeutic Targets Database [87][39] (TTD, [88]http://xin.cz3.nus.edu.sg/group/cjttd/ttd.asp, Version 4.3.02, released on Aug 25th 2011) to screen drugs similar to DB via structural similarity comparisons. We only selected drugs with a high structural similarity score (>0.85, similar to very similar) with the potentially absorbed DB constituents and their metabolites. The therapeutic targets of these similar drugs were also included as putative DB targets. 7.2. Network construction and analysis The absorbed DB components and their metabolites, their putative targets, and known therapeutic targets for colitis treatment were used to construct a chemical component-putative target network, and a putative DB targets-known colitis therapeutic targets PPI network, respectively. PPI data were obtained from eight existing PPI databases, as described in section 2.6. Navigator software (Version 2.2.1) and Cytoscape (Version 2.8.1) were utilized to visualize the networks. For each node “i” in these two networks, we defined four measures of topology: (1) “degree,” defined as the number of links to node i; (2) “betweenness,” defined as the capacity of node i to be located in the shortest communication paths between different pairs of nodes in the network [89][40]. The betweenness centrality of node i is computed as following formula: graphic file with name pone.0101432.e001.jpg where Inline graphic denotes the number of the shortest paths between node s and node t in the PPI network, Inline graphic denotes the number of the shortest paths across node i between node s and node t, and N is the total number of nodes in the PPI network. This property correlates more closely with essentiality than connectivity, exposing critical nodes that usually belong to the group of scaffold proteins or proteins involved in crosstalk between signal pathways (called bottlenecks) [90][41]. (3) “closeness,” defined as the inverse of the sum of node i distances to all other nodes. Closeness can be regarded as a measure of how long it would take to spread information from node i to all other nodes sequentially. Degree, betweenness and closeness centralities correlate with a protein's topological importance in the PPI network [91][40]. (4) K-core analysis is an iterative process in which nodes were removed from the network in order of the least-connected [92][42]. The core of maximum order is defined as the main (highest) k-core in the network. A k-core sub-network of the original network can be generated by recursively deleting vertices from the network whose degree is less than k. This results in a series of sub-networks that gradually reveal the globally central region of the original network. On this basis, “k value” is used to measure the centrality of node i. 7.3 Pathway enrichment analysis for the major putative DB targets and the major known therapeutic targets for colitis We performed pathway enrichment analysis using pathway data obtained from the FTP service of KEGG [93][43] (Kyoto Encyclopedia of Genes and Genomes, [94]http://www.genome.jp/kegg/, Last updated: Oct 16, 2012). Results and Discussion 1. Optimization of UPLC-ESI-MS/MS conditions To obtain reliable chromatographic results and appropriate ionization, four mobile phase systems of acetonitrile-water, methanol-water, acetonitrile-acid aqueous solution, and methanol-acid aqueous solution were tested and compared. The results suggested that acetonitrile-formic acid aqueous solution was superior to the others. Different concentration of formic acid aqueous solution (0.05%, 0.1%, and 0.5%) were investigated, indicating that use of 0.1% formic acid aqueous solution produced the optimal peak shape and reduced peak tailing, as well as increasing ion response for most compounds. Meanwhile, the process of gradient elution was optimized to obtain better separation over shorter time periods. The MS parameters were optimized to improve ion intensity. These optimized parameters were described in section 2.3. Base peak chromatograms of DB in positive and negative ion modes are shown in [95]Figure 2 . Figure 2. UPLC-ESI-MS/MS of DB in positive ion mode (A) and negative ion mode (B). [96]Figure 2 [97]Open in a new tab 2. Screening and identification of DB using UPLC-ESI-MS/MS Recently, the UPLC-Q-Exactive system was reported to provide a rapid, reliable and accurate technique for identification of herbal constitutions, because of its separation efficiency, high resolution and high mass accuracy [98][44]–[99][46]. In the present study, a total of 48 components of DB were identified. These were mainly flavonoids, such as flavones, flavonols, flavanones, chalcones, and isoflavones, as listed in [100]Table 1 . For some compounds (7,4′-dihydroxyflavan, loureirin A, loureirin B, pterostilbene, and dracaenin A), purified standards were available and these were identified by comparing sample retention time and accurate mass with that of the standard. For compounds where standards were unavailable, identification was based on accurate mass and tandem mass spectra. The molecular formula was established by high-accuracy quasi-molecular ion such as [M-H]^−, [M+CH[3]COO]^−, [M+H]^+, and [M+Na]^+ within a mass error of 5.0 ppm and fractional isotope abundance. The most rational molecular formula was identified using chemical or herbal databases, such as Chemspider ([101]www.chemspider.com) and TCM Database@Taiwan ([102]http://tcm.cmu.edu.tw). When several isomers were matched, a compound that had previously been identified in DB was considered more likely to be correct. Finally, ion fragments were used to provide further confirmation of the chemical structure. The structures of the main constitutions of DB are shown in [103]Figure 3 . For example, 7,4′-dihydroxyflavan (compound 5) was firstly identified by comparing its retention time and accurate mass with that of the appropriate standard. Then, the MS/MS spectrum and possible fragmentation pathways of 7,4′-dihydroxyflavan were depicted in [104]Figure 4 . Only [M+H]^+ and [M–H]^− were observed in positive and negative ion modes, respectively. The loss of 128 Da and 136 Da were the characteristic neutral loss of 7,4′-dihydroxyflavan in positive and negative ion modes. Table 1. MS data of ESI-MS spectra and identification of the Longxuejie enteric-coated tablet. No. RT (Measured) m/z (Measured) m/z (Expected) m/z (Delta, ppm) polarity Formula Compound Name Papp (10^−6 cm/s) 1 8.11 311.11395 311.11363 1.046 negative C[15]H[20]O[7] 3,4-dihydroxyallylbenzene 4-O—D-Glucopyranoside 3.8494 2 8.38 285.07592 285.07575 0.594 positive C[16]H[12]O[5] unknown — 283.06137 283.0612 0.606 negative 3 8.68 229.08608 229.08592 0.702 positive C[14]H[12]O[3] 3,5,4'-trihydroxystilbene 148.9186 227.07154 227.07137 0.752 negative 4 8.68 273.07713 273.07685 1.054 negative C[15]H[14]O[5] 4'-methoxy-3'7-dihydroxyflavan 115.6472 5 8.72 255.06525 255.06519 0.241 positive C[15]H[10]O[4] 7,4'-dihydroxyflavan 137.4393 253.05074 253.05063 0.408 negative 6 8.74 259.09663 259.09649 0.563 positive C[15]H[14]O[4] 2,4,4'-Trihydroxydihydrochalcone 94.5322 257.08211 257.08193 0.707 negative 7 8.83 314.13891 314.13868 0.703 positive C[18]H[19]NO[4] norisoboldine 30.448 312.12448 312.12413 1.11 negative 8 9.27 149.06084 149.0608 0.219 negative C[9]H[10]O[2] unknown — 9 9.66 269.08209 269.08193 0.594 negative C[16]H[14]O[4] unknown — 10 9.78 525.19107 525.19078 0.558 positive C[32]H[28]O[7] dracaenin A 134.7937 523.1766 523.17623 0.709 negative 11 10.02 273.11223 273.11214 0.359 positive C[16]H[16]O[4] unknown — 271.09769 271.09758 0.385 negative 12 10.21 303.1228 303.1227 0.339 positive C[17]H[18]O[5] unknown — 301.10831 301.10815 0.531 negative 13 10.21 392.20704 392.20676 0.696 positive C[21]H[26]O[6] 3,4-O-dimethylcedrusin 149.4172 14 10.21 167.0704 167.07027 0.788 positive C[9]H[10]O[3] Ethyl 4-hydroxybenzoate 216.6625 15 10.57 287.09152 287.0914 0.414 positive C[16]H[14]O[5] unknown — 285.07703 285.07685 0.638 negative 16 10.73 532.23343 532.23298 0.849 positive C[31]H[30]O[7] cochinchinenin 21.059 513.19218 513.19188 0.587 negative 17 10.95 301.10837 301.10815 0.733 negative C[17]H[18]O[5] unknown — 18 11.2 241.08608 241.08592 0.665 positive C[15]H[12]O[3] (2S)-7-hydroxyflavanone 227.3156 239.07151 239.07137 0.616 negative 19 11.45 199.07646 199.07645 0.057 negative C[13]H[12]O[2] 4,4'-Dihydroxydiphenylmethane 201.7499 20 11.64 529.22251 529.22208 0.817 positive C[32]H[32]O[7] unknown — 527.20758 527.20753 0.102 negative 21 11.65 511.21352 511.21261 1.776 negative C[32]H[32]O[6] homoisosocotrin-4-ol 17.1673 22 11.86 283.09773 283.09758 0.507 negative C[17]H[16]O[4] unknown — 23 12 516.23823 516.23806 0.321 positive C[31]H[30]O[6] socotrin 4-ol 7.5670 24 12.13 287.12781 287.12779 0.077 positive C[17]H[18]O[4] Loureirin A 228.3831 285.11337 285.11323 0.492 negative 25 12.13 151.07542 151.07536 0.442 positive C[9]H[10]O[2] 3,4-Dihydroxy-allylbenzene 140.4007 26 12.23 227.10674 227.10666 0.351 positive C[15]H[14]O[2] 7-hydroxyflavanone 227.3156 27 12.23 317.13839 317.13835 0.136 positive C[18]H[20]O[5] Loureirin B 227.8266 315.12406 315.1238 0.833 negative 28 12.93 257.11728 257.11722 0.227 positive C[16]H[16]O[3] pterostilbene 236.2931 255.10277 255.10267 0.396 negative 29 13 885.45033 885.44894 1.572 negative C[44]H[70]O[18] 26-O—D-Glucopynanosyl-furostan-5,25(27)-diene-1,3,22,26-tetrol-1-O-[-L- rhamnopyranosyl(12)]—L-Arabinopyranoside 0.0073 30 13.25 167.0704 167.07027 0.799 positive C[9]H[10]O[3] unknown — 31 14.08 265.14798 negative unknown — 32 14.39 462.32139 462.3214 −0.029 positive C[27]H[40]O[5] dracaenol B 141.9713 33 14.90 571.23383 negative unknown — 34 14.91 527.24323 527.24282 0.786 positive C[33]H[34]O[6] cochinchinenin E 8.2809 35 15.77 532.92554 negative unknown — 36 15.92 325.18451 325.18569 −3.642 negative C[14]H[29]O[8] unknown — 37 16.27 415.32104 415.32067 0.894 positive C[27]H[42]O[3] Diosgenin 112.3202 38 16.44 325.18448 325.18569 −3.734 negative C[14]H[29]O[8] unknown — 39 17.44 339.20010 339.20134 −3.668 negative C[15]H[31]O[8] unknown — 40 19.18 698.91071 negative unknown — 41 19.95 323.25817 323.25807 0.292 positive C[20]H[34]O[3] bincatriol 172.7858 42 20.25 445.29987 negative unknown — 43 20.49 628.19489 positive unknown — 44 20.63 383.18988 383.19117 −4.807 negative C[16]H[31]O[10] unknown — 45 21 461.32938 positive unknown — 46 21.6 116.92857 negative unknown — 47 22.14 394.3465 394.34415 4.577 Positive C[25] H[46] O[3] unknown — 48 24.99 427.3939 427.39344 1.082 positive C[30]H[50]O lupeol 0.0386 [105]Open in a new tab Papp: apparent permeability coefficient at the indicated concentrations in the basolateral-to-apical (B-A) direction across a Caco-2 cell monolayer. Figure 3. The structures of the 20 potentially absorbed DB constituents and 8 metabolites. [106]Figure 3 [107]Open in a new tab 3. 3,5,4′-trihydroxystilbene; 4. 4′-methoxy-3′7-dihydroxyflavan; 5. 7,4′-dihydroxyflavan; 6. 2,4,4′-trihydroxydihydrochalcone; 7. norisoboldine; 10. dracaenin A; 13. 3,4-o-dimethylcedrusin; 14. ethyl 4-hydroxybenzoate; 16. cochinchinenin; 18. (2S)-7-hydroxyflavanone; 19. 4,4′-dihydroxydiphenylmethane; 21. homoisosocotrin-4-ol; 24. loureirin A; 25. 3,4-dihydroxy-allylbenzene; 26. 7-hydroxyflavanone; 27. loureirin B; 28. pterostilbene; 32. dracaenol B; 34. cochinchinenin E; 37. diosgenin; 41. bincatriol. M1 and M2 were metabolites of compound 3; M3, M4, M5, M6, M7, and M8 were metabolites of compounds 4, 5, 25, 27, 28, and 37, respectively. Figure 4. The MS/MS spectrum and possible fragmentation pathways of 7,4′-dihydroxyflavan. [108]Figure 4 [109]Open in a new tab 3. Prediction of absorbed DB constituents and their metabolites by in silico ADME models Quantitative structure-permeability relationships (QSPerR) have been established for permeability across Caco-2 monolayers (Papp) by the application of new molecular descriptors [110][47]. Among them, the passive Caco-2 absorption model provides a useful drug discovery tool by predicting human oral absorption of new chemical entities [111][48]–[112][49]. Yazdanian et al. [113][50] reported that compounds with Papp values less than 0.4×10^−6 cm/s exhibited very poor oral absorption, whereas compounds with Papp values greater than 7×10^–6 cm/s had excellent oral absorption. The present study employed an in silico passive Caco-2 absorption model within the ACD/Percepta software to identify DB constituents with excellent oral absorption values (greater than 7×10^−6 cm/s) and the results were shown in [114]Table 1 . Subsequently, the ACD/Percepta software P450 regioselectivity module was applied to predict the metabolites of these well-absorbed DB constituents. These results were listed in [115]Table 2 and the structures of the metabolites were shown in [116]Figure 3 . Table 2. Prediction of the score, reliability, reaction site and reaction type of excellent oral absorbed constituents of LXJ by the P450 regioselectivity module in ACD/Percepta, respectively. No. Compound No Compound name Score Reliability Reaction site Reaction type Metabolizing enzymes M1 3 3,5,4'-trihydroxyStilbene 0.94 0.86 3 Aromatic Hydroxylation CYP1A2 0.94 0.86 7 Aromatic Hydroxylation CYP1A2 M2 3 3,5,4'-trihydroxyStilbene 0.65 0.87 14 Aromatic Hydroxylation CYP1A2 M3 4 4'-methoxy-3'7-dihydroxyflavan 0.92 0.78 21 O-dealkylation CYP1A2 M4 5 7,4'-Dihydroxyflavan 0.92 0.79 13 Aromatic Hydroxylation CYP1A2 0.92 0.79 15 Aromatic Hydroxylation CYP1A2 M5 25 3,4-Dihydroxy-allylbenzene 0.85 0.74 4 Aliphatic Hydroxylation CYP2C19 0.91 0.86 4 Aliphatic Hydroxylation CYP2C9 M6 27 Loureirin B 0.74 0.5 18 Aromatic Hydroxylation CYP1A2 0.74 0.5 19 Aromatic Hydroxylation CYP1A2 M7 28 pterostilbene 0.90 0.74 10 Aromatic Hydroxylation CYP1A2 0.90 0.74 13 Aromatic Hydroxylation CYP1A2 M8 37 Diosgenin 0.64 0.52 24 Aliphatic Hydroxylation CYP3A4 [117]Open in a new tab 4. Analysis of DB pharmacological mechanisms in colitis In order to analyze the synergistic effects and pharmacological mechanisms of DB in colitis, we constructed a chemical component-putative target network, and a putative DB targets-known colitis therapeutic targets PPI network. Detailed information relating to these similar drugs and putative targets is provided in supplementary [118]Table S3. The performance of this prediction method was evaluated in our previous study [119][34]. 4.1. DB chemical component-putative target network As shown in [120]Figure 5 , this network consisted of 67 nodes (15 absorbed components, 4 metabolites and 48 putative targets) and 120 edges. The mean number of putative targets per chemical component/metabolite was 2.53. The absorbed DB component 14 and the metabolite M5 of compound 25 (3,4-dihydroxy-allylbenzene) both had the highest degree distributions, and hit 23 and 14 putative targets, respectively. Since chemical components or metabolites with a higher degree in the network have been demonstrated to be more pharmacologically important, and our data indicated that components 25, 32, and the M5 metabolite had the shortest paths between each other, they may play major roles in this pharmacological network, with synergistic effects. Figure 5. DB chemical component-putative target network. [121]Figure 5 [122]Open in a new tab Blue squares refer to the DB compounds and metabolites; yellow spherical nodes refer to the predicted targets. 4.2. Putative DB targets-known colitis therapeutic targets PPI network Putative DB targets-known colitis therapeutic targets PPI network was constructed using PPIs between putative DB targets and known colitis therapeutic targets. As shown in [123]Figure 6A , this network consisted of 2397 nodes and 4005 edges. According to the previous study of Li et al. [124][51], if the degree of a node is more than 2 fold of the median degree of all nodes in a network, such node is believed to play a critical role in the network structure, and can be treated as a hub node. Thus, we identified 463 hub proteins. Then, the PPI network of these hub proteins was constructed using the direct PPIs between them. As a result, this network consisted of 463 nodes and 1574 edges ([125] Figure 6B ). In addition, we calculated four topological feature values of all 463 nodes in the PPI network of hub proteins according to the definition described in section 2.7.2. We defined a hub protein, the four topological feature values of which were all larger than the corresponding median values, as a major hub protein. After the calculation, the median values of the four topological features were 3, 0.06, 31.30, and 3 for “degree,” “betweenness,” “closeness,” and “k value,” respectively. Therefore, we determined that hub proteins with “degree” >3, “betweenness” >0.06, “closeness” >31.30, and “k value” >3 were major hubs. As a result, 106 proteins were identified as major hubs, including 18 putative DB targets and 16 known therapeutic targets for colitis. Detailed information on the topological features of these major hubs is provided in supplementary [126]Table S4. Figure 6. Putative DB targets-known colitis therapeutic targets protein-protein interaction (PPI) network. [127]Figure 6 [128]Open in a new tab (A) The network between all targets and other human proteins. (B) The network of hub proteins in network (A). (C) The network of the major putative DB targets and the major known colitis therapeutic targets in network (B). Yellow spherical nodes indicate the putative targets; pink spherical nodes indicate the known therapeutic targets; purple spherical nodes indicate other human proteins that interact with putative targets or known therapeutic targets. Red edges in (C) indicate the PPIs of targets involved in the NOD-like receptor signaling pathway. To investigate the pharmacological mechanisms of DB in colitis further, we analyzed the direct PPIs between the major putative DB targets and the major known therapeutic targets for colitis. As shown in [129]Figure 6C , this network contained 34 nodes and 42 edges. According to the results of enrichment analysis based on the KEGG pathway [130][43], these major target proteins were frequently involved in the NOD-like receptor signaling pathway (KEGG ID: map04621). The intracellular NOD-like receptor (NLR) family contains more than 20 members in mammals and plays a pivotal role in the recognition of intracellular ligands. NOD1 and NOD2, two prototypic NLRs, sense the cytosolic presence of bacterial peptidoglycan fragments that have escaped from endosomal compartments, driving the activation of NF-kappa-B and MAPK, cytokine production, and apoptosis [131][52]. Accumulating studies have demonstrated that several members of the NOD-like receptor signaling pathway, such as heat shock protein 90 (Hsp90) [132][53], interleukin-6 (IL-6) [133][54], tumor necrosis factor alpha (TNF-α) [134][55], NF-kappa-B inhibitor kinase alpha (IKKA) [135][56], IKKB [136][56], and adrenergic receptors [137][57], may be associated with the pathogenesis of colitis. Of these proteins, Hsp90, ADRB1, and ADRB2 were identified as putative DB targets. Moreover, a number of putative DB targets also interacted with members of the NOD-like receptor signaling pathway, as shown in [138]Fig. 6C . These findings suggested that the therapeutic effects of DB on colitis may involve this pathway. Conclusion High-throughput analysis, in silico ADME prediction, and network pharmacology techniques have recently emerged as powerful tools to provide new insights into the active compounds and the molecular mechanisms involved in the therapeutic actions of traditional Chinese medicines. In this study, a rapid, sensitive, and high-throughput analytical method using UPLC-ESI-MS/MS was firstly developed to characterize 48 constituents and accurately identify 24 constituents of DB tablet. Then, 22 components of DB tablet were predicted to be absorbed by an in silico passive absorption model, based on the Caco-2 cell monolayer. Eight metabolites were predicted using a computational module of P450 regioselectivity. Finally, these compounds were predicted to interact with 26 putative targets. Using this information, pharmacological networks were built to visualize the synergistic interactions between DB components and their targets. Furthermore, a PPI network of putative DB targets and known therapeutic targets for colitis was constructed to pinpoint the key targets and pathways. This approach offered a valuable opportunity to deepen understanding of the pharmacological mechanisms of DB in colitis. In summary, this research provided a more accurate analysis of DB constituents, compared to using TCM databases. Consideration of ADME using the passive absorption and P450 metabolism modules in silico predicted the in vivo situation more closely. Finally, network pharmacology analysis was applied effectively and helped to interpret the essence of “synergy” in Chinese medicine. This provided a new way to identify the key candidate targets and possible molecular pathways utilized by DB in the treatment of colitis. Supporting Information Table S1 The detailed target information for colitis. (XLSX) [139]Click here for additional data file.^ (87.2KB, xlsx) Table S2 The detailed PPI information of colitis based on eight existing PPI databases. (XLSX) [140]Click here for additional data file.^ (9.8KB, xlsx) Table S3 Detailed information relating to these similar drugs and putative targets for DB curing colitis. (XLSX) [141]Click here for additional data file.^ (61.6KB, xlsx) Table S4 Detailed information on the topological features of these major hubs including 18 putative DB targets and 16 known therapeutic targets for colitis. (XLSX) [142]Click here for additional data file.^ (13.5KB, xlsx) Data Availability The authors confirm that all data underlying the findings are fully available without restriction. Relevant data are included within the paper and its Supporting Information files. Funding Statement Financial support was provided by Key Projects in the National Science & Technology Pillar Program during the 11th Five-Year Plan Period (No. 2008BAI51B03), the National Natural Science Foundation of China (No. 81202793), Beijing Joint Project Specific Funds and independent topics supported by Operational expenses for basic research of China Academy of Chinese Medical Sciences (No. Z02063), Beijing Joint Project Specific Funds and the Fundamental Research Funds for the Central public welfare research institutes (Nos. ZZ070819, ZZ070831 and ZZ070830). Support was also provided in the form of salary for Guangrong Zhu as an employee at Datang Hanfang Pharmacy co.ltd. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. References