In silico network pharmacology-based dissection of mechanisms of the Chinese herb Radix Isatidis as an effective treatment of respiratory tract diseases.
With a history of more than 2500 years of empirical testing and refinement, Traditional Chinese Medicine (TCM) is regarded as a valuable treasure for Asia. It aims to make dysfunctioning living organisms return to their normal states in a holistic way that has already played a significant role in the maintenance of health among Asians (Wang, 2012). Elsewhere around the globe, it is common to use TCM as well. It is reported that the United States, who is the largest importer of TCM products from China, spent as much as US$7.6 billion alone in 2015. Given the importance of TCM, more and more attention has been drawn onto the relating study. However, the working mechanism of TCM which relates to the therapeutic effectiveness is still unknown in most circumstances due to the lack of scientific and technological approaches (Wang, 2012). Since multiple active components in the herbs may exert synergistic therapeutic efficacies, diversified ingredients in TCM as well as a wide range of related targets complicate the pharmacological research significantly, making it difficult to uncover the mysterious mechanisms using traditional experimental methods. Consequently, a comprehensive approach which is able to capture the holistic property of TCM is really needed in research.
As a newly emerged field of pharmacology, in silico network pharmacology utilizes network analysis to help understand the mechanism of multiple actions of drugs across multiple scales ranging from molecular and cellular levels to tissue and organism levels. Based on genomics, proteomics, metabolomics and some other technological platforms, it enables the study of the essence of TCM and the function of herbal ingredients in a holistic way. Furthermore, a significant area of integration between in silico network pharmacology and drug discovery is polypharmacology which describes a phenomenon that many effective drugs act on more than one target while these targets involve multiple diseases. By mapping the in silico polypharmacology network, both the explicit targets hit by active compounds and additional diseases related with important drug targets might be uncovered (Hopkins, 2007; Abreu, 2015)). Thus the application of in silico network pharmacology to TCM provides new possibilities to understand the interactions among active ingredients of herbs, relevant targets and various diseases which in turn highlight the mechanism of action. Presently, we will use a previously developed robust model based on in silico network pharmacology to shed light on the mystery of one of the oldest TCMs, namely, Radix Isatidis (Li, 2012).
Radix Isatidis (Indigowoad root, Isatis root, or Banlangen in Chinese) is a very famous traditional Chinese medicine (TCM) extracted from the roots of Isatis indigotica Fort. Its broad antiviral activity has been used for thousands of years in China. This herb is cultivated in various regions of China and harvested for its root during the annual autumn season. The sliced root is processed into granules or syrups, and then dissolved or diluted in hot water. Radix Isatidis and its finished products have important functions in preventing and treating influenza, tonsillitis, and malignant infectious diseases, especially severe acute respiratory syndrome (SARS) and H1N1-influenza because of its anti-viral, anti-bacterial, anti-inflammatory, anti-tumor, and immune regulatory functions. Moreover, numerous gratifying successes of the Radix Isatidis anti-viral effect have been reported. Radix Isatidis has become an important component in various traditional Chinese medicine preparations, of which Banlangen Granules is widely used for respiratory tract diseases in clinical practice. However, several fundamental questions still remain uncovered including: 1) what the active ingredients of the herb are; 2) what explicit targets the active ingredients hit; and 3) how these components exert effects on various diseases.
In the present work, an integrated computer model of in silico network pharmacology has been used to cope with the above questions (Li, 2012). Developed in our previous study, this model combines druglikeness evaluation, oral bioavailability prediction, multiple drug targets prediction as well as network pharmacology techniques. We firstly identified its candidate compounds (more likely to be the active ingredients) and potential targets, and then mapped these compounds and targets onto functional ontologies to generate several drug-target-disease networks which help to understand the holistic and synergic essence of the herbal medicine from a systematic point of view. The obtained results may not only provide new insights for a deeper understanding of the chemical and pharmacological basis of Radix Isatidis, but also provide an efficient way for drug discovery from herbal medicine.
2. Materials and methods
2.1. Database building
After discarding those compounds with unknown structures, we finally obtained a total of 177 compounds as Radix Isatidis's ingredients from our own database: TCMSP: Traditional Chinese Medicines for Systems Pharmacology Database and Analysis Platform (http://lsp.nwsuaf. edu.cn/tcmsp.php). As a chemically oriented herbal encyclopedia, TCMSP is capable of providing detailed and up-to-date information about herbal components and structural data. These ingredients were saved as mol2 format for further analysis and were optimized by Sybyl 6.9 (Tripos Associates, St. Louis, MO) with the same parameters as described in our previous work (Zhang, 2011). Considering the process of an orally administrated Chinese herb, the glycosyls of the ingredients will be deglycosylated by enteric bacteria in the intestine. As a result, 10 components of Radix Isatidis with glycosyls were deglycosylated according to the rule of glycosidase hydrolysis reaction. The obtained products were further optimized by procedures as mentioned above.
2.2. Drug-likeness (DL) evaluation
Given the costs, length and complexity of the drug-discovery process, it is vital to recognize what a drug is like. In order to obtain drug-like compounds, a database-dependent model was applied to calculate the drug-likeness (see equation) of each compound by using the Tanimoto coefficient (Thorner, 1997), where 'a' is the molecular property of each compound while 'b' represents the average molecular properties of the whole compounds in the Drugbank database (http://www. drugbank.ca/).
DL(a,b) = ab/[[parallel]a[parallel].sup.2] + [[parallel]b[parallel].sup.2] - ab
The Drugbank database is a unique bioinformatics and cheminformatics resource that contains data of 6511 compounds. These compounds are either FDA approved drugs or chemicals under clinical trials which have a great possibility to become licensed drugs. When calculating their average drug-likeness index, a value of 0.18 was obtained (Liu, 2013). Thus we assume if one compound has drug-likeness larger than 0.18, it is more likely to be developed into a drug in the future. Therefore, DL[greater than or equal to]0.18 is regarded as one threshold for screening possible candidate drugs presently.
2.3. Oral bioavailability (OB) prediction
Oral bioavailability, one of the most important pharmacokinetic parameters, represents the speed of a drug to become available to the body and the eventually absorbed extent of the oral dose (Turner, 2004). Since poor OB is indeed a main reason responsible for the unsuccessful development of compounds into therapeutic drugs in drug screening cascades, it is valuable to conduct oral bioavailability screening on the compounds. However, the prediction of OB is very challenging because of the complex function of many biological and physicochemical factors affecting bioavailability. However, thanks to the rapid development of computational chemistry, we have developed a novel and robust in-house software OBioavailm to predict the oral bioavailability for each compound (Xu, 2012). This software was built based upon 805 structurally different drugs and drug-like molecules and was integrated with the transport (P-glycoprotein) and metabolism (P450 3A4) information. According to the previous standard developed by Wang et al., OB[greater than or equal to]30% was selected as another screening threshold and the qualified molecules satisfying both DL and OB thresholds were regarded as candidate compounds. The OB criterion used here mainly focuses on two principles: (1) Using the fewest and most suitable active ingredients to extract as much information as possible from Radix Isatidis; (2) The reported pharmacological data will give a reasonable explanation of the obtained model.
2.4. Target prediction
Considering the fact that most drugs act by binding to specific proteins and some Traditional Chinese Medicines might target multiple proteins due to the existence of its multiple active components, target identification is helpful to elucidate the mechanism of action of Radix Isatidis through the analysis of drug-target and target-disease networks. In the present work, based on a systematic model developed in our previous work, we predicted the candidate targets by using candidate ingredients of the herb (Yu, 2012). This model was built based upon 6511 drugs and 3999 targets with known compound-protein interactions in the Drugbank database (http: //www. drugbank.ca/). The chemical, genomic and pharmacological information for drug targeting was integrated on a large scale and two powerful mathematical methods namely Random Forest (RF) and Support Vector Machine (SVM) were applied. As an effective tool in property prediction, RF is a classification and regression algorithm (Amaratunga, 2008). It can be developed as a classifier which consists of a number of decision trees while each tree casts a unit vote for the most popular class (Amaratunga, 2008). When dealing with the prediction of drug-target interactions, RF is more robust against the overfitting problem and runs more efficiently on large dimensional datasets (Yu, 2012). SVM is another promising classification and regression method, which maps the input vectors to a very high- non-linear dimensional feature space, to get a maximal margin hyperplane. Because of its good performance in dealing with linear non-separable problems, SVM has also been applied to predict the interaction between drugs and proteins (Yu, 2012). Presently, the Random Forest software package (http://www.stat.berkeley.edu/users/breiman/) and the LIBSVM suite of programs (http://www.csie.ntu.edu.tw/cjlin/libsvm) were used to build the RF and SVM prediction models, respectively. The performance of the predicted models is impressive with a concordance of 85.83%, a sensitivity of 79.62%, a specificity of 84.12% for the RF optimal model and a concordance of 70.80%, a sensitivity of 41.78%, and specificity of 92.76% for the SVM optimal model. In this work, the overlap targets between the RF and SVM models located in the top 100 ranking targets were treated as candidate targets. These targets were further validated by the Uniprot (http://www.uniprot.org/), Drugbank (http://www.drugbank. ca/) and TTD (http://bidd.nus.edu.sg/group/ttd/). Taking the Uniprot as an example, one mission of Uniprot is to provide a comprehensive, high-quality and freely accessible resource of functional information about proteins. The name of each candidate target was inputted into the search box of the Uniprot database to get its functional annotation. Then we would check if its functional information was relevant with antiviral activities or regulation of the immune system.
2.5. Drug-target-disease network construction
To elucidate the multicomponent therapeutic mechanism of Radix Isatidis in the treatment of respiratory tract diseases from a network target perspective, we constructed two visualized networks: candidate compound-potential target network (Compound-Target network) and potential target-disease network (Target-Disease network), to understand the holistic and synergic essence of herbal medicine from a systematic point of view. The Compound-Target network was constructed by linking the candidate compounds to the corresponding validated potential targets while the Target-Disease network was built by linking the potential targets to their relevant diseases. Information on the disease was mined by mapping the terminology of the potential targets to related disease in the TTD database. The obtained 129 diseases were then classified into 13 disease categories under Medical Subject Headings (http://www.nlm.nih.gov). Both networks were generated by Cytoscape 3.2.1, an open source bioinformatic package for biological network analysis and visualization (Smoot, 2011). In this graphic network, a node represents a candidate compound, a target or a disease, while an edge encodes the interaction of candidate compound-potential target or potential target-disease. In order to disclose the importance of a node and how this node affects the signal transmission of related nodes, two statistical parameters namely degree and betweenness were applied to the analysis of the obtained network. The former is defined as the number of which a node connects to, while the latter is the ratio of the number of shortest paths passing through a node to the number of total paths passing through the nodes.
3. Results and discussion
Traditional Chinese Medicine, developed through thousands of years of observation and accumulated experience, poses a stable therapeutic effect in a more holistic way when compared with western medicine (Wang, 2012). However, due to the great number of different biological components contained in each herbal medicine as well as the complex background theories of TCM, the therapeutic mechanism is still ambiguous for most of the medicinal plants in the TCM research. Radix Isatidis, one of the traditionally used herbal medicines, has been widely used as an anti-virus and immune enhancing medicine. In our work, a novel in silico pharmacology approach integrated with chemogenomics, polypharmacology and network biology was applied to uncover the therapeutic mechanism of Radix Isatidis from a systematic level.
3.1. Drug-likeness and oral bioavailability evaluation
Drug-likeness (DL), a qualitative property of chemicals, refers to how pharmacokinetic and pharmaceutical properties of compounds like solubility and potency correspond to the majority of known drugs, in brief how "drug like" a prospective chemical is (Fukunishi, 2011). Considering the value of drug-likeness on saving time and costs in the process of drug discovery, it was used to filter out compounds with undesirable pharmaceutical properties in the database (Wang, 2012). In addition, apart from drug-likeness, oral bioavailability (OB) is also an important pharmacokinetic parameter to screen candidate compounds which have the potential to be further developed into drugs. Oral bioavailability (OB) represents the percentage of an oral dose that reaches the systemic circulation and produces a pharmacological effect (Li, 2012). It has been reported that most chemicals in the TCM fail to reach the active site and produce a pharmacological effect because of their poor pharmacokinetic properties, especially the OB (Li, 2012). Consequently, predicting the OB of the known compounds of Radix Isatidis is necessary in order to obtain candidate compounds that play key roles in the pharmacodynamic process.
The obtained result shows that 39 candidate compounds were screened from the 172 chemicals (22.7%) which possess not only satisfactory drug-likeness ([greater than or equal to]0.18), but also OB ([greater than or equal to] 30.0%). Interestingly, it has been proved that some of the candidate compounds possess good pharmacological effects on hypertension or the regulation of the immune system. For instance, Linarin (MOL001790, DL=0.71, OB=39.8%), 7,4'-Dihydroxyflavanone (MOL001792, DL=0.18, OB=32.8%), Neohesperidin_qt (MOL001798, DL=0.27, OB=71.2%), Saponaretin (MOL002322, DL=0.72, OB=31.3%) and Indirubin (MOL002309, DL=0.26, OB=48.6%) were reported to exhibit anti-viral activity. Indigo Blue (MOL001781, DL=0.26, OB=38.2%) possesses very clear antibacterial activity, and indirubin (MOL002309, DL=0.26, OB=48.6%) which also has a good effect in the treatment of leukemia for its structure-activity relationships, thus improving the preventive effect on colon cancer by inhibiting the tumor cell's DNA synthesis.
In order to avoid omitting active ingredients, we also reviewed abundant related articles to manually add a few literature-based active ingredients. For example, Syringinenin (MOL000347, DL=0.32, 0B=14.6%), Sinigrin (MOL013312, DL=0.18, OB=3.9%), Goitrin (MOL001811, DL=0.01, OB=3.2%), Sinigrin_qt (MOL001761, DL=0.03, OB=21.2%) being the major bioactive ingredients of Radix Isatidis. exhibits anti-virus, anti-mocrobial, antiendotoxin, and anti-cancer effects, although its DL or OB index is extremely low. These four compounds are manually added into the active ingredients database.
It is worth noting that Neohesperidin (MOL013383, DL=0.69), one of the most abundant ingredients in Radix Isatidis, shows a relatively poor OB value of 11.6% even though this compound is able to present effective anti-virus, anti-endotoxin and anti-mocrobial activities. This can be explained by the previous study that the oral absorption of neohesperidin is relatively low and most of the Neohesperidin has been biotransformed into various metabolites by the microbe in the colon. As a result, this compound should be taken into account in the following discussion. In addition, After observing the obtained data and reading related published papers carefully, we found that 3-(2-hydroxyphenyl)-4- quinazolinone (MOL001763, DL=0.i6, Ob=65.6%), 4-Quinazolone (MOL001764, DL=0.04, OB=39.i%), Isoliquiritigenin (MOL001789, DL=0.15, OB=85.3%), Syringic acid (MOL001807, DL=0.06, OB=47.8%), Benzoic acid (MOL000219, DL=0.03, OB=60.4%) and Salicylic acid (MOL001801, DL=0.03, OB=32.1%) present either potent anti-virus activity or the ability to promote the immune system, with relatively high OB but poor DL. In addition, according to the report from University of Wisconsin, Tryptantherin (MOL001808, DL=0.29, 0B=19.3%) used as an antibacterial agent could kill plaster-like ringworm fungus, purple ringworm fungus, red ringworm fungus, microsporum lanosum, microsporum gypseum, and epidermophytonfloccosum, so it identified as the major antimicrobial principle of Radix Isatidis. Thus, the eight ingredients of Radix Isatidis were also regarded as candidate compounds. To sum it up, 51 ingredients of Radix Isatidis antiviral activities or regulation of the immune system are finally screened and regarded as candidate compounds.
3.2. Target identification
Since TCM contains a lot of pharmacologically effective compounds, a pivotal challenge lies in the identification of molecular targets which will help to shed light on the mechanism of TCM's action in a system-level view. Traditionally, the sequencing of expressed-sequence tags (ESTs), serial analysis of gene expression (SAGE) differential display, homology cloning and relevant approaches have been used to confirm the therapeutic targets of drugs through experiments. However, given the time-consuming, expensive, challenging feature and a narrow application scope of this process, computational methods should be the first choice. Thanks to the application of our systematic model based on the RF and SVM methods, presently the binding of candidate components of Radix Isatidis to those targets that are related with antiviral activities and/or the enhancement of immune system have been studied.
The 51 candidate compounds yield 112 candidate targets and the connections between them reach up to 794. The obtained targets were further subjected to the Uniprot, Drugbank and TTD databases to check if they are relevant with respiratory tract diseases. Finally, 29 potential targets (Table 1) are reserved and 6 candidate compounds (MOL001726, MOL001728, MOL001769, MOL001811, MOL013312, MOL013383) without any relevant targets are removed, indicating that most of the candidate compounds we screened previously have effective therapeutic actions. Furthermore, the average target hit by one compound has been calculated and ended up with a value of 1.333, showing that the action of some compounds may be promiscuous.
3.3. Compound-target-disease network analysis
The well accepted "one drug, one target" theory has been found to be less effective than hoped because of the intrinsic robustness of living systems against various erturbations. Consequently, the focus of drug discovery has shifted to the "multi-drug, multi-target" theory. Interestingly, given the multiple ingredients of TCM and its remarkable pharmacological efficacy in the body, this theory may be able to help uncover the synergistic effect between multi-components and multi-targets. However, due to the method used in the TCM research that focuses mainly on the path of partitioned reductive analysis, the characteristic of the scientific system of a herbal medicine is unable to be captured and the synergistic therapeutic mechanism is still poorly uncovered (Liu, 2013). Thanks to the rapid development of network pharmacology recently, the above problems can be solved from a system point of view. Presently, the network method is applied to detect both the candidate ingredients of Radix Isatidis, and the effective targets interacted with these ingredients as well as to elaborate the mechanism of action of this herbal medicine.
3.3.1. Compoud-Target network: detecting key players of Radix Isatidis for treating diseases
After deleting 6 compounds with no targets, the remaining 45 candidate compounds and their relevant 29 potential targets were adopted to generate a bipartite graph of drug-target interaction network. Figure. 1 shows a global view of Compoud-Target network in which blue purple circle and pink arrow represent the candidate compounds and potential targets of Radix Isatidis, respectively. This network consists of 77 nodes (45 candidate compounds and 29 potential targets) and 292 edges. The centralization and heterogeneity of this net are 0.361 and 0.867 respectively, indicating that certain compounds or targets are more central and biased than the others in the network. For example, prostaglandin G/H synthase 2 (T18) has the largest number of drug interactions while aldose reductase (T47) has only one drug interaction.
[FIGURE 1 OMITTED]
As a fundamental topological parameter of network, degree may pinpoint and offer insights into highly influential compounds or targets (Azuaje, 2011). Interestingly, the degree of some nodes displays a rather big number of compound-target interactions while others are middle or small. This is consistent with the previous results of centralization and heterogeneity (Table 2). Beta-sitosterol (MOL000358) exhibits the largest number of target connections (18), followed by Stigmasterol (MOL000449) with 15 potential targets and Acacetin (MOL001689), Tryptantherin (MOL001808) and Indirubin (MOL002309) with 13 potential targets. Similarly, the more central potential targets are prostaglandin G/H synthase 2 (T18), dipeptidyl peptidase IV (T53) and prostaglandin G/H synthase 1 (T2) connected by 32, 29 and 28 candidate compounds, respectively. These high-degree nodes are referred to as hubs (Azuaje, 2011), suggesting their more important role in helping to treat an influenza virus infection or enhance the immune system. Interestingly, of all the 45 candidate compounds, 15 compounds possess degree larger than 10 under an average value of 7.5 in which 8 are reported to be active compounds. The excellent hit rate (>50%) suggests the reliability and rationality of our network analysis to pinpoint the key players of Radix Isatidis.
Another basic parameter of network nodes is betweenness, a key measure to assess the relevance of the location of nodes within a network. A node is considered central if it involves lots of paths linking pairs of nodes. Interestingly, we find that values of degree and betweenness are correlated with each other strongly and the node with high betweenness tends to possess large degree. For example, the top ten targets with high betweenness values are prostaglandin G/H synthase 2 (T18), dipeptidyl peptidase IV (T53), cell division protein kinase 2 (T67), glucocorticoid receptor (T51), prostaglandin G/H synthase 1 (T2), heat shock protein HSP 90 (T64), glycogen synthase kinase-3 beta (T61), DNA topoisomerase II (T46), peroxisome proliferator activated receptor gammareceptor (T13) and beta-lactamase (T69), that are exactly the same as the top ten targets with large degree except for the slight difference of target order. Consequently, compounds or targets with higher degree and betweenness would be key players of Radix Isatidis.
3.3.2. Compoumd-Target Network: illustrating the mechanism of action of Radix Isatidis on treating diseases based on the compound-target interactions
Based on holism, a range of potential targets may be modulated by multiple compounds from TCM in a synergistic manner and a new equilibrium of living system finally achieved with less harmful impact. This "multi-drug, multi-target" theory is probably the therapeutic mechanism of Radix Isatidis.
1. Inflammation: Inflammation is a fundamental protective response that enables human survival when encountering a microbial invasion or injury, and also maintains the tissue homeostasis under various deleterious surroundings. The deficiency of inflammatory responses always results in immunodeficiency phenomena such as cancer and infection, while the superfluity of inflammatory responses often contributes to higher incidence rate and mortality in several diseases. As a natural plant for the treatment of inflammation, Radix Isatidis has been widely used in Asia for a long time . Among all 45 potential targets some of them have a high probability of being able to actively target inflammation, these are: prostaglandin G/H synthase 1 (T2), peroxisome proliferator activated receptor (T13), prostaglandin G/H synthase 2 (T18), Nitric-oxide synthase, endothelial (T19), beta-2 adrenergic receptor (T39), tumor necrosis factor (T44), mitogen-activated protein kinase 14 (T58), and leukotriene A-4 hydrolase (T70). These inflammatory relevant targets may synergistically mediate the living system in various tissues and organs and make the system return to a healthy state after getting hit by a series of compounds.
2. Respiratory tract diseases: After a careful inspection of our network, we find that a number of targets related to a large number of compounds that could possibly be developed to treat respiratory tract diseases, these include: Radix Isatidis, including Prostaglandin G/H synthase 1 (T2), peroxisome proliferator activated receptor gammareceptor (T13), arachidonate 5-lipoxygenase (T17), deltatype opioid receptor (T27),cGMP-inhibited 3',5'-cyclic phosphodiesterase A (T29), muscarinic acetylcholine receptor M2 (T37), beta-2 adrenergic receptor (T39), tumor necrosis factor (T44), mu-type opioid receptor (T49), mitogen-activated protein kinase 14 (T58), transcription factor AP-1 (T60) and E-selectin (T62).
3.3.3. Target-disease network
Given the fact that most complex diseases are triggered by an unbalanced modulating network in which multiple genes or their products stay in a state of dysfunction and that proteins targeted by a drug tend to associate with various diseases, we have constructed a Target-Disease network to explore other potential therapeutic effects of Radix Isatidis. In this network, each potential target has been found to have its relevant diseases and all 29 targets achieved 129 diseases that were classified into 13 groups according to the Medical Subject Headings (http://www.nlm.nih.gov). For example: Asthma, Adult respiratory distress syndrome, and inflammatory lung disease belong to Respiratory tract diseases, while inflammatory bowel disease, diarrhea and ulcerative colitis fall under Digestive system diseases. Figure.2 shows a global view of Target-Disease network in which pink arrow nodes and blue diamond nodes represent the potential targets and relevant diseases, respectively.
[FIGURE 2 OMITTED]
Among the 129 diseases, we find that most of them belong to respiratory tract diseases (24/129), inflammatory diseases (10/129), Infectious and parasitic diseases (15/129), neoplasms (76/129), mental disorders (13/129), Cardiovascular Diseases (20/129), nervous system diseases (12/129), indicating the potential therapeutic role of Radix Isatidis in the treatment of these diseases. For instance, transcription factor AP-1 (T60) which is predicted to be the potential target of Beta-sitosterol (MOL000358) may be a therapeutic target for the treatment of a series of diseases, including breast cancer (neoplasms), allergic airway inflammation (respiratory tract diseases) and rheumatoid arthritis (musculoskeletal diseases). It has been reported that beta-sitosterol could stimulate transcriptional activity of estrogen responsive elements in the promoters of target genes through the modulation of estrogen receptor beta, indicating the possible therapeutic effect of this compound in curing the above mentioned diseases. These results confirm the good performance of our Target-Disease network and provide a clear view of the relationships between candidate ingredients of the herb and the diseases through the protein targets.
Presently, an in silico network pharmacology methodology integrating; drug-likeness evaluation, oral bioavailability prediction, multiple drug targets prediction and network analysis has been applied to the investigation of the therapeutic mechanism of Radix Isatidis. Our main findings are as follows:
1. 51 ingredients of Radix Isatidis were screened and are regarded as candidate compounds. 29 Potential targets hit by these ingredients were identified.
2. The network analysis clearly elucidates the synergistic action mechanism of Radix Isatidis on the treatment of respiratory tract diseases based on the "multidrug, multi-target" theory about compound-target interactions.
3. The Target-Disease network displays that Radix Isatidis has certain therapeutic efficiency for the control of other diseases such as endocrine system diseases, neoplasms, mental disorders, nervous system diseases, cardiovascular diseases, digestive system diseases, skin and connective tissue diseases, musculoskeletal diseases and genitourinary system diseases, indicating that the medicinal herb may also be used to treat multiple diseases.
4. The present work provides an alternative in silico strategy for the deep investigation and understanding of the chemical and pharmacological basis of TCM, which will promote drug discovery from herbal medicines.
Thanks for the support given by the high-performance computing platform of Northwest A&F University and funding of national college students' innovation and entrepreneurship training program (Grant no. 201510716494).
Amaratunga, D., Cabrera, J., Lee, Y.S. (2008). Enriched random forests. Bioinformatics, 24 (18), 2010-2014.
Azuaje, F. J., Zhang, L., Devaux, Y., Wagner, D. R. (2011). Drug-target network in myocardial infarction reveals multiple side effects of unrelated drugs. Scientific reports, 1, 52.
Abreu, A., Rocha, A., Cota, M. P., & Carvalho, J. V. (2015). Caderneta Eletronica no Processo Ensino-Aprendizagem: Visao de Professores e Pais de alunos do ensino Basico e Secundario. RISTI--Revista Iberica de Sistemas e Tecnologias de Informacao, (16), 108-128.
Fukunishi, Y., Nakamura, H. (2011). Definition of drug-likeness for compound affinity. Journal of chemical information and modeling, 51(5), 1012-1016.
Hopkins, A. L. (2007). Network pharmacology. Nature biotechnology, 25(10), 1110-1111.
Kitano, H. (2007). A robustness-based approach to systems-oriented drug design. Nature reviews. Drug discovery, 6 (3), 202-210.
Li, X., Xu, X., Wang, J. (2012). A system-level investigation into the mechanisms of Chinese Traditional Medicine: Compound Danshen Formula for cardiovascular disease treatment. PloS one, 7 (9), 43-91.
Liu, H., Wang, J., Zhou, W., Wang, Y., Yang, L. (2013). Systems approaches and polypharmacology for drug discovery from herbal medicines: an example using licorice. Journal of ethnopharmacology, 146 (3), 773-793.
Smoot, M. E., Ono, K., Ruscheinski, Wang, P. L., Ideker, T. (2011). Cytoscape 2.8: new features for data integration and network visualization. Bioinformatics, 27(3), 431-432.
Thorner, D. A., Willett, P., Wright, P. M., Taylor, R. (1997). Similarity searching in files of three-dimensional chemical structures: representation and searching of molecular electrostatic potentials using field-graphs. Journal of computer-aided molecular design, 11(2), 163-174.
Turner, J. V., Maddalena, D.J., Agatonovic-Kustrin, S. (2004). Bioavailability prediction based on molecular structure for a diverse series of drugs. Pharmaceutical research, 21(1), 68-82.
Wang, X., Xu, X., Tao, W. (2012). A systems biology approach to uncovering pharmacological synergy in herbal medicines with applications to cardiovascular disease. Evidence-based complementary and alternative medicine: eCAM, 5, 19-31.
Xu, X., Zhang, W., Huang, C. (2012). A novel chemometric method for the prediction of human oral bioavailability. International journal of molecular sciences, 13(6), 6964-6982.
Yu, H., Chen, J., Xu, X. (2012). A systematic prediction of multiple drug-target interactions from chemical, genomic, and pharmacological data. PloS one, 7(5), e37608.
Zhang, H.X., Li, Y., Wang, X. (2011). Insight into the structural requirements of benzothiadiazine scaffold-based derivatives as hepatitis C virus NS5B polymerase inhibitors using 3D-QSAR, molecular docking and molecular dynamics. Current medicinal chemistry, 18 (26), 4019-4028.
Hailong Shi (1,2), Yaya Cui (3), Yucheng Wang (1), Yingying Fan (1), Jiaxin Gong (1), Hong Zhang (1), Lin Dang (1), Shuguang Yan (1), Xinrong Guo 4*
(1) College of Basic Medicine, Shaanxi University of Chinese Medicine, Xi'an-Xianyang New Ecomic Zone, Shaanxi Province, 712046, China
(2) Laboratory for Functional Glycomics, College of Life Sciences, Northwest University, Xi'an City, Shaanxi Province, 710069, China
(3) Office of External Cooperation and Development, Shaanxi University of Chinese Medicine, Xi'an-Xianyang New Ecomic Zone, Shaanxi Province, 712046, China
(4) Department of Acupuncture and Moxibustion, Shaanxi University of Chinese Medicine, Xi'an-Xianyang New Ecomic Zone, Xi'an City, Shaanxi Province, 712046, China
Table 1--The information of 29 potential targets No. Potential target T2 Prostaglandin G/H synthase 1 T17 Arachidonate 5-lipoxygenase T25 Muscarinic acetylcholine receptor M4 T37 Muscarinic acetylcholine receptor M2 T31 Cathepsin K T46 DNA topoisomerase II T51 Glucocorticoid receptor T58 Mitogen-activated protein kinase 14 T62 E-selectin T69 Beta-lactamase T76 Cellular tumor antigen p53 T11 Sodium channel protein type 5 subunit alpha T18 Prostaglandin G/H synthase 2 T27 Delta-type opioid receptor T39 Beta-2 adrenergic receptor T33 cGMP-inhibited 3',5'-cyclic phosphodiesterase A T47 Aldose reductase T53 Dipeptidyl peptidase IV T60 Transcription factor AP-1 T64 Heat shock protein HSP 90 T70 Leukotriene A-4 hydrolase T82 Glutathione S-transferase P T13 Peroxisome proliferator activated receptor gammareceptor T19 Nitric-oxide synthase, endothelial T29 CGMP-inhibited 3,5-cyclic phosphodiesterase A T27 cAMP-specific 3',5'-cyclic phosphodiesterase 4D T44 Tumor necrosis factor T49 Mu-type opioid receptor T57 Fatty acid synthase T61 Glycogen synthase kinase-3 beta T67 Cell division protein kinase 2 T73 Neuronal acetylcholine receptor protein, alpha-7 chain Table 2--The degree and betweenness of candidate compounds and potential targets in network Node Degree Betweenness MOL000219 3 0.000417960 MOL000347 11 0.020875280 MOL000358 18 0.150949890 MOL000359 1 0.000000000 MOL000449 15 0.102226720 MOL000953 2 0.004718590 MOL001398 5 0.002336170 MOL001689 13 0.059977420 MOL001721 8 0.024522350 MOL001722 8 0.003247670 MOL001733 11 0.025942300 MOL001734 6 0.001920880 MOL001735 10 0.007717440 T11 13 0.029121000 T13 20 0.034297590 T17 1 0.000000000 T18 32 0.186662390 T19 8 0.011218480 T2 28 0.088088700 T25 2 0.000579230 T27 2 0.000236900 T29 7 0.003566280 T37 3 0.000978180 MOL001736 9 0.003989040 MOL001749 2 0.000135070 MOL001755 1 0.000000000 MOL001756 4 0.000660850 MOL001763 12 0.021489870 MOL001764 2 0.000429220 MOL001767 8 0.002678410 MOL001774 1 0.000000000 MOL001779 10 0.042715880 MOL001781 3 0.000277150 MOL001782 7 0.002038530 MOL001783 5 0.007525540 MOL001789 11 0.036409050 T39 11 0.025237580 T44 1 0.000000000 T46 6 0.034585010 T47 1 0.000000000 T49 2 0.000744330 T51 6 0.091065440 T53 29 0.111727080 T57 2 0.000762580 T58 19 0.024602030 T60 1 0.000000000 MOL001790 1 0.000000000 MOL001792 10 0.006558300 MOL001793 7 0.002038530 MOL001798 10 0.006361740 MOL001801 7 0.067899390 MOL001803 8 0.016525490 MOL001808 13 0.018549220 MOL001814 10 0.007443140 MOL001820 12 0.013856120 MOL001828 9 0.004698740 MOL001833 3 0.000470160 MOL002309 13 0.018549220 MOL002322 3 0.032075730 T61 22 0.039265470 T62 1 0.000000000 T64 23 0.043733490 T67 27 0.108438960 T69 14 0.031390490 T70 3 0.000625470 T73 6 0.002815520 T76 1 0.000000000 T82 1 0.000000000