PDTCM: a systems pharmacology platform of traditional Chinese medicine for psoriasis

Abstract

Introduction: Psoriasis is a refractory skin disorder, and usually requires a lifetime control. Traditional Chinese medicine (TCM) is effective and safe for this disease. However, the cellular and molecular mechanisms of TCM remedies for psoriasis are still not fully understood. TCM contains numerous natural products. Natural products have historically been invaluable as a resource of therapeutic agents. Yet, there is no integrated information about active compounds of TCM for psoriasis.

Method: We use systems pharmacology methods to develop the Psoriasis Database of Traditional Chinese Medicine (PDTCM). The database covered a number of psoriasis-related information (formulas, TCM, compounds, target proteins, diseases and biomarkers). With these data information, an online platform was constructed

Results: PDTCM comprises 38 empirical therapeutic formulas, 34373 compounds from 1424 medicinal plants, 44 psoriasis-related proteins and 76 biomarkers from 111 related diseases. On this platform, users can screen active compounds for a psoriasis-related target and explore molecular mechanisms of TCM. Accordingly, users can also download the retrieved structures and data information with a defined value set. In addition, it helps to get a better understanding of Chinese prescriptions in disease treatment.

Conclusion: With the systems pharmacology-based data, PDTCM would become a valuable resource for TCM in psoriasis-related research.

Key messages

• PDTCM platform comprises a great deal of data on TCM and psoriasis.

• On this platform, users can retrieve and get needed information with systems pharmacology methods, such as active compounds screening, target prediction and molecular mechanisms exploration.

• It is a tool for psoriasis-related research on natural drugs systematically.

Keywords:

• High throughput virtual screening
• psoriasis
• systems pharmacology
• target prediction
• traditional Chinese medicine

1. Introduction

Psoriasis is a chronic recurrent cutaneous disorder and usually persists for life [1,2]. Psoriatic lesions are frequently itchy or painful, or both. It is probably the most prevalent immune-mediated skin disease in adults and affects approximately 2% of people worldwide [3,4]. At least 10% of patients developed irreversible damage to articular cartilage [5]. Psoriasis suffers report an impairment of their quality of life, because of the unsightly appearance of erythematous scaly skin lesions [6].

Natural products (NPs) are the important source of drugs [7,8]. For example, methotrexate, psoralen plus ultraviolet A, cyclosporin and acitretin are from a NP or a modified NP, which are widely used for systemic therapeutic options [9]. However, they were not highly satisfactory [10]. Patients using biologic agents reported satisfaction with their treatment, but biologics cost much more than other systemic remedies and also require repeated injections [11,12]. Evidence from systematic reviews has shown the efficacy of traditional Chinese medicines (TCM) in psoriasis treatments [13,14], and TCM is a popular complementary therapy in western countries [15–17]. It is well known that TCM is all from nature, containing abundant natural products. NPs are a valuable resource of new drug candidates. As the quality of leads arising from NP discovery is better and more biologically friendly [18], scientists strongly suggested that expanding the exploration of NP as a crucial source of novel active agents for disease indications [19–21].

Psoriasis is a worldwide refractory disease. Our team has been focusing on herbal treatment for psoriasis over the years. On clinical and basic research, we found that TCM formula could reduce recurrence rate and regulate the pathogenesis pathways [22,23]. Inspired by the significant sources of approved drugs derived from NPs in drug discovery, we designed and established the Psoriasis Database of Traditional Chinese Medicine (PDTCM) with systems pharmacology-based methods, such as high throughput virtual screening and network analysis. The database comprised 34373 compounds from plants, animals and microorganism. To our knowledge, PDTCM is the first non-commercial and freely available database for natural psoriatic drugs (http://cadd.gdhtcm.com:2180/PDTCM/). On this platform, we compared TCM and approved clinical psoriatic drugs, and try to explain mechanisms of natural drugs in psoriatic treatment from the view of complex networks in systems pharmacology. It can also provide clues and ideas in exploring perspective and effective TCM for psoriasis.

2. Materials and methods

2.1. Data collections

PDTCM comprises eight data entities including herbal formulas, TCM, source plants/animals, molecules, target proteins, docking results between all molecules and target proteins, diseases and clinical biomarkers. The herbal formulas were selected from classical literature of TCM and monographs by senior famous TCM physicians, which were used for treating “Wan xuan”, “Bai bi”, “Niu pi xuan” and “Yin xie bing”. TCM was collected by keywords involved in psoriatic features, such as inflammation or inflammatory skin diseases, dermatitis, immune, angiogenesis, keratosis and proliferation. The TCM’s latin names were retrieved from Chinese Materia Medica and Chinese Pharmacopoeia in parts of indications, pharmacological effects or clinical applications. The components of TCM were collected from Reaxys, Chinese Natural Product Database [24], Traditional Chinese Medicines Database [25] and Chinese Herbal Drug Dasabase [26]. The 3D structures of compounds were obtained from the Universal Natural Products Database (UNPD) [27]. Targets of psoriasis and approved drugs were retrieved from DrugBank (http://www.drugbank.ca) [28], Uniprot (http://www.uniprot.org) and TTD (http://bidd.nus.edu.sg/BIDD-Databases/TTD/TTD.asp) [29]. The information of diseases, clinical biomarkers related to psoriasis targets and their relationship with target proteins were collected from KEGG [30], OMIM (http://www.ncbi.nlm.nih.gov/omim).

2.2. Chemical space and drug-likeness analysis

Molecular descriptors of the molecules in PDTCM and 10 approved drugs coming from NPs and modified NPs (calcipotriol, clobetasol, salicyclic acid, etretinate, tazarotene, fluticasone propionate, desoximetasone, alitretinoin, acitretin and calcitriol) were calculated by Discovery Studio by using default parameters (Table 1). Eight descriptors: lipo-hydro partition coefficient (AlogP), molecular weight (MW), molecular fractional polar surface area, number of hydrogen bond donors (NHBD), number of hydrogen bond acceptors (NHBA), number of rotatable bonds, number of rings and number of aromatic rings were selected to build the chemical space. Principal component analysis (PCA) was conducted in the library analysis module of Discovery Studio to visualize the distributions of molecules and drugs in the chemical space. PCA between natural compounds and approved drugs was performed in two-dimension space.

Table 1. Statistics of molecular descriptors in PDTCM.

	Compounds			Drug
Descriptors	Mean ± StdDev	Min	Max	Mean ± StdDev	Min	Max
ALogP	2.50 ± 3.24	−18.7	39.85	4.39 ± 1.48	1.22	6.24
Molecular_Weight	488.37 ± 274.94	30.07	3740.53	364.42 ± 95.53	138.12	500.57
Num_RotatableBonds	6.4 ± 6.1	0	66	4.9 ± 1.9	1	8
Num_Rings	4.1 ± 2.5	0	32	2.6 ± 1.4	1	4
Num_AromaticRings	1.0 ± 1.3	0	20	0.5 ± 0.7	0	2
Num_H_Acceptors	8.4 ± 7.2	0	104	3.6 ± 1.1	2	6
Num_H_Donors	4.1 ± 4.2	0	57	1.4 ± 1.0	0	3
Molecular_Fractional_Polar_Surface_Area	0.27 ± 0.13	0	0.91	0.18 ± 0.09	0.09	0.42

2.3. Molecular docking

Molecular docking is used as a standard computational tool in structure-based drug design [31]. In order to screen potential active compounds, molecular docking was employed. The crystal or NMR structures of the target proteins listed in Table 2 were downloaded from the RCSB Protein Data Bank (http://www.rcsb.org/pdb/home/home.do). The protein–ligand complex structures of target proteins were dealt with by removing the hetero atoms and adding hydrogen atoms. Then, molecular docking was performed by Autodock 4 in DOVIS 2.0 [32]. For each target protein, the binding site was defined as a 40 × 40 × 40 Å cube centred on the occupied space of the original ligand with a spacing of 0.375 Å between the grid points. Based on the docking results, we constructed the molecule-target network (MTN) and calculated properties of nodes by Cytoscape 3.0.2 [33].

Table 2. Targets related to psoriasis for molecular docking.

Uniprot ID	PDB entry	Uniprot ID	PDB entry	Uniprot ID	PDB entry	Uniprot ID	PDB entry
O60674	3E64	P12318	3RY6	P28482	3I5Z	P52333	3LXL
P00374	1MVT	P13631	1EXX	P28702	1UHL	P53778	1CM8
P00533	1XKK	P14780	2OVX	P29460	1F42	P62942	2DG3
P01375	1A8M	P16581	1G1T	P31151	1PK6	Q02750	3DY7
P04150	1NHZ	P17948	3HNG	P35228	2NSI	Q07343	1XLX
P05106	2VDM	P19793	1LBD	P35354	3LN1	Q08499	3G4K
P06702	1IRJ	P20292	2Q7R	P35968	2QU5	Q08881	1A8M
P08253	1HOV	P20701	1RD4	P36897	2X7O	Q13946	1CWM
P10276	3KMR	P23219	3N8X	P39900	1JK3	Q14289	3FZS
P10826	1XDK	P23458	1DB1	P42345	1AUE	Q15759	3GC9
P11473	1DB1	P27815	3I8V	P47712	1CJY	Q16539	1KV1

All structures are coming from PDB database (http://www.rcsb.org/pdb/home/).

2.4. Lead compound discovery

Virtual screening (VS) is a computational approach to evaluate very large libraries of compounds, which has become an integral part in the process of drug discovery [34]. With a manageable number of compounds filtered by VS, it helps to improve the efficiency of lead compound discovery and reduce the cost of experimental test in vitro and in vivo. Molecule docking is a method of VS based on the target structure. Docking scores are used to measure the binding affinity between molecules and targets [35,36]. Accordingly, those with high scores should be selected to further analysis and to be tested.

2.5. Case study: clinical prescriptions in Chinese Pharmacopoeia

Chinese Pharmacopoeia is an authoritative work in clinical and practical application. Fufang Qingdai Wan, Yujin Yinxie Pian and Xiaoyin Pian are the only three prescriptions recorded in Chinese Pharmacopoeia for psoriasis treatment (Supplementary Table S1). Therefore, we chose the three typical formulas as an example of usage. First, we retrieved the composed molecules and molecule-target interactions of each formula from PDTCM. With the extracted information, we plotted three formulas’ PCA in three-dimensional system, and constructed the MTNs, respectively.

3. Results and discussion

3.1. Content and utility

A total of 38 empirical formulas and 1424 TCM have been collected into PDTCM. In addition, the database comprises 3D structures of 34,373 compounds from 1424 TCM, and the interactions between compounds and target proteins. As a result, formulas, TCM, compounds, targets, target-related diseases and disease biomarkers are linked together on this platform. PDTCM has a concise and intuitive web interface equipped with five sections (Home, Introduction, Browse, Search and Download), and users can get information by screen taxonomy capturing. A retrieval system for molecules is available on the homepage. Users can query for exact compounds by the followings: (i) molecule ID, (ii) CAS number, (iii) chemical name, (iv) molecular formula, (v) canonical SMILES, (vi) InChIKey and (vii) natural source. The introduction page presents contents and functions of database. The browse section displays all of the empirical formulas, source plants, TCM, compounds, target proteins and diseases by classification. On the search part, users can get interactions between herbs, target proteins and diseases. Specifically, with a designed search module of herbal formula DIY, users can predict the material basis and active mechanisms of a herbal combination flexibly. For instance, to identify candidate drug targets and active compounds of a psoriasis therapeutic formula “PSORI-CM01” [37], the steps are shown in Figure 1. By submitting the parameters, formula’s target proteins and physicochemical properties of screened compounds were displayed, and structures could be downloaded by link.

Figure 1. An illustrated steps for classic formula “PSORI-CM01” (composed of seven herbs). Notes: It was achieved on the page of Herbal formula DIY.

3.2. Molecular properties and chemical space analysis between TCM and approved drugs for psoriasis

Molecular descriptors have been used to display structural diversity of a set of compounds with transformed chemical information [38]. The chemical space contributed by molecular descriptors encompassed all possible small organic molecules [39]. Wide distribution of molecules in the chemical space would enhance the opportunities to discover leads through similar occupancy with pharmaceuticals [40]. Table 1 lists the molecular descriptors of 34,373 compounds from 1424 TCM and psoriatic drugs. Obviously, statistical means and standard deviations of TCM were larger than those of approved drugs. Consequently, TCM would provide a basis for modulation of different biological targets, which was important for pharmaceutical product development [41].

PCA can convert structural data into a new coordinate system by an orthogonal linear transformation method. PCA revealed the internal structural features between various classes of compounds in a graphic with the transformed original variables [42]. To visualize the in chemical space, we plotted PCA. The data set for PCA comprised compounds from 1424 TCM and 10 psoriatic drugs (Figure 2). Figure 2 displays a distinct visualization of distributed features of molecules in PDTCM and psoriatic drugs. The overlapped part in the PC1–PC2 projection suggested that the compounds in PDTCM would have a number of drug-like compounds with similar structures of psoriatic drugs.

Figure 2. Distribution in chemical space according to PCA of compounds in PDTCM and psoriatic drugs.

The drug-like space has been defined using cut-off values for permissible physical properties [43–45]. Till now, Lipinski’s rule of five (Ro5) has been widely used as a classical methods to predict drug-like molecules, although some effective drugs do not obey Ro5 [46]. The percentage of molecular properties in TCM that obey Ro5 is present in Figure 3. Figure 3 shows more than a half of molecules met the range of each drug-like property, which implied TCM might have good oral bioavailability.

Figure 3. Distribution of drug-like molecular descriptors of TCM.

3.3. Lead compound discovery by virtual screening

TCM contains a vast number of NPs with biological activities [47]. With the development of computer technology, high throughput virtual screening could be used to predict NPs’ biological activities [48]. In order to promote the accuracy of predicted results and lower the complex degree of data handling, the threshold we set was that the docking score of a molecule was above 8.0, and higher than that of target’s original ligand. With the criterion we set, the docking results were extracted and MTN was constructed. Figure 4 displays the phenomenon of a compound targeting several proteins, and some compounds acting on a target. In detail, 643 nodes (605 molecules and 38 proteins) and 844 edges were constructed in the network (Supplementary Table S2). In MTN, a compound targeted an average of 1.4 target proteins and each target protein had 22.2 hits. Degree and betweenness centrality, both two parameters implied the pivotal nodes in a network. These nodes would be noteworthy for potential polypharmacologic compounds. Table 3 presents molecules with high degree and betweenness centrality, such as M010178, M018782 and M026498. M010178 and M026498 were separated from Hypericum perforatum. M018782 was a compound isolated from Strychnos nux-vomica. Hypericum perforatum was confirmed for anti-inflammatory effects, and the extract could modulate the immune system by inhibiting tumour necrosis factor-alpha [49,50]. Studies had reported the biological activities of Sophora alopecuroides, Smilax china and S. nux-vomica in inflammatory skin diseases [51–53]. For further validation, the compounds listed in Table 3 should be further investigated.

Figure 4. Drug-target network of PDTCM. Notes: Triangle and dot represent molecules (n = 605) and targets (n = 38), respectively.

Table 3. Top rank of potential compounds for psoriasis according to degree and betweenness centrality of MTN.

Rank	Compound	CAS NO.	Degree	Betweenness centrality	Chemical name	Source
1	M010178	N/A	9	0.04	Hypericinate	Hypericum perforatum
2	M018782	N/A	7	0.11	Nb-methyl-longicaudatine	Strychnos nux-vomica
3	M026498	N/A	7	0.02	Pseudohypericinate	Hypericum perforatum
4	M031574	N/A	7	0.01	N/A	Hypericum perforatum
5	M003603	N/A	6	0.01	L-arginine	Smilax china
6	M012471	7034-04-0	6	0.01	Tricrotonyltetramin	Sophora alopecuroides
7	M020429	1422351-81-2	6	0.01	Strychnoflavine	Strychnos nux-vomica
8	M025067	54352-30-6	6	0	Base A	Sophora alopecuroides
9	M002659	548-04-9	5	0.03	Hypericin	Hypericum perforatum
10	M007335	N/A	5	0.01	Isopseudohypericin	Hypericum perforatum
11	M008571	72393-03-4	5	0.01	Fagopyrin	Fagopyrum esculentum
12	M019628	1421445-06-8	5	0.01	Demethoxyguiaflavine	Strychnos nux-vomica

3.4. Network pharmacology for clinical prescriptions

Traditional Chinese medicine focuses on the whole body and the prescription varies on the syndrome instead of a disease. It has been stated in Chinese Pharmacopoeia, that Fufang Qingdai Wan (FFQDW) can clear heat and reduce toxicity, which is used for progressive stage of psoriasis or blood-heat syndrome in TCM. Yujin Yinxie Pian (YJYXP) can promote blood circulation, which could be used for blood stasis syndrome. Xiaoyin Pian (XYP) can nourish blood and dispel wind, which is used for blood dryness syndrome. Constituent compounds and duplicate herbs of these formulas are listed in Table 4. Apparently only two ingredient herbs were the same at most between the two formulas. However, based on the graph of PCA, the large overlap indicated the existence of compounds with similar structures in the three formulas (Figure 5). The overlap part suggested the consistency of all the three formulas for treating psoriasis according to structural features. The scattered parts might be aiming at psoriasis different syndromes. In order to explore molecular mechanisms of three clinical formulas, we constructed the interactions networks between their molecules and targets separately. Network analysis indicated that arachidonate 5-lipoxygenase-activating protein (ALOX5AP) was their collaborative target. ALOX5AP was required for arachidonic acid metabolite synthesis. Accordingly, arachidonic acid metabolism participated by ALOX5AP would be the three prescriptions’ common pathway for anti-inflammatory regulation.

Figure 5. PCA of three clinical formulas.

Table 4. Three formulas information about composed compounds and duplicate herbs.

Formulas	Duplicate herbs	Num. of compounds
FFQDW	Baphicacanthus cusia; Dictamnus dasycarpus	884
YJYXP	Angelica sinensis; Carthamus tinctorius; Baphicacanthus cusia	1014
XYP	Angelica sinensis; Carthamus tinctorius; Dictamnus dasycarpus	1339

4. Conclusion and future directions

PDTCM is the first database that embraces a wide range of TCM, and systems pharmacology-based data between compounds and psoriatic target proteins. PDTCM has been designed not only to provide a platform for psoriatic lead compound discovery, but also to provide a better understanding of the molecular mechanisms of Chinese prescriptions. As a result, it helps to broad the formula’s indications and the application of TCM in disease treatment. In the course of new confirmed targets, PDTCM website data will be integrated and updated continuously. Future efforts will also replenish effective prescriptions from formal clinical trials, as well as the structures of newly discovered compounds from natural medicines. Based on the provided information, such as predicted compounds and refined formulas, experimental studies are still needed for users to validate the predictions. Overall, PDTCM is intended to be a long-term resource for users to obtain systems pharmacology data on psoriasis-related TCM.

Acknowledgements

The authors thank the whole team for assistance and fruitful discussions.

Disclosure statement

The authors report no conflicts of interest.

Additional information

Funding

National Key Technology R&D Program for the 12th Five-year Plan of Ministry of Science and Technology, China2013BAI02B03

Natural Science Foundation of Guangdong Province10.13039/501100003453S2013030011515

National Natural Science Foundation of China10.13039/50110000180981302736

Guangdong Science and Technology Department10.13039/5011000071622014A0202210402016A0202260102017A020213011

Financial Industry Technology Research Development Program of Guangdong Province of China2011(285)05

Guangdong Provincial Hospital of Chinese MedicineYK2013B1N02YN2014ZH07

Guangzhou University of Chinese MedicineA1-AFD018161Z04

Guangdong Science and Technology project2013B090700015

the 12th Five-year Plan of Ministry of Science and Technology, China2013BAI02B03

This work was supported by the National Key Technology R&D Program for the 12th Five-year Plan of Ministry of Science and Technology, China (2013BAI02B03); team project of Natural Science Foundation of Guangdong Province (S2013030011515); National Natural Science Foundation of China (81302736); Guangdong Science and Technology Department (2014A020221040, 2016A020226010, 2017A020213011); the Financial Industry Technology Research Development Program of Guangdong Province of China [2011(285)05]; special research on Chinese Medical Science and Technology of Guangdong Provincial Hospital of Chinese Medicine (YK2013B1N02, YN2014ZH07); and the start-up support for scientific research of Xinglin Yong Scholar in Guangzhou University of Chinese Medicine (A1-AFD018161Z04).