SerpentinaDB: a database of plant-derived molecules of Rauvolfia serpentina

In order to build an extensive repository of PDMs from R. serpentina, data were compiled from literature and web resources. All resources were manually curated to extract PDMs data and their additional details including plant part, chemical name, chemical class, and IUPAC name. To address degeneracy in the name of the plant, PubMed (http://​www.​ncbi.​nlm.​nih.​gov/​pubmed) was searched with two variants of spelling (‘Rauvolfia serpentina’ and ‘Rauwolfia serpentina’) to obtain relevant information. A total of 31 research articles, 3 books, 2 PhD dissertations[13, 14], and 3 web resources involving natural compounds research were utilized to compile an extensive list of PDMs. Books and web resources used for curation of dataincluded following sources: ‘The Alkaloids’ [15], ‘The Alkaloids: Chemistry and Physiology’ [16], ‘The Alkaloids: Chemistry and Physiology’ [17], A database on antidiabetic plants [18], Global Information Hub On Integrated Medicine [19], and India Herbs [20]. To authenticate the chemical details obtained, molecules were also ascertained from the Dictionary of Natural Products (DNP) [21], PubChem (https://​pubchem.​ncbi.​nlm.​nih.​gov/​) [22], ChemSpider (http://​www.​chemspider.​com/​) [23], and ChEMBL (https://​www.​ebi.​ac.​uk/​chembl/​) [24]. 3D chemical structures of molecules were drawn and edited using MarvinSketchv5.10.0 software (https://​www.​chemaxon.​com), and further saved into mol2 file format. These files were subjected to energy minimization with Merck Molecular Force Field (MMFF94) using OpenBabel v2.3.1 software [25]. Mol2 format is desirable due to ease of conversion into other molecular formats accepted by drug discovery softwares. To remove redundant entries, data from all resources were merged, and an extensive library of PDMs was compiled. A total of 147 molecules, reported to be extracted from various plant parts, were present in final dataset (Fig. 2a) and broadly classified into different chemical classes (Fig. 2b). Representative PDM of each of seven chemical classes is provided in the supplementary file (Additional file 3). A separate entry was created for molecules that were obtained from more than one plant part leading to 227 such individual entries. Of all these PDMs, mol2 files for 5 of them could not be obtained due to unavailability of both structure as well as IUPAC. Physicochemical properties of the PDMs, such energy, atomic contribution to the partition coefficient (AlogP), distribution coefficient (logD), molecular formula, molecular mass, molecular solubility, and molecular weight (MW), acid dissolution constant (pKa), number of aromatic bonds, number of aromatic rings, and radius of gyration, hydrogen bond acceptor (HBA) count, hydrogen bond donor (HBD) count, number of H acceptor, number of H donor, number of H acceptor (Lipinski), number of H donor (Lipinski), number of H bonds, solvent accessible surface area (SASA), and surface area as well as ADMET (ADMET Solubility, ADMET Solubility Level, ADMET BBB, and ADMET BBB Level) properties, were obtained using molecular property finder tool under small molecules category from the Discovery Studio v4.0 (Accelrys, San Diego, USA). Figure 3 illustrates the statistics of various physicochemical properties, such as MW (Fig. 3a), HBA as well as HBD (Fig. 3b), and molecular volume (MV) (Fig. 3c).

SerpentinaDB is hosted in a Server at the Indian Institute of Technology Jodhpur on a Dell Power Edge R910 server running a Linux operating system (Red Hat version 5.5). A total of seven data tables were created to house compiled data. SerpentinaDB implements MySQL, an object-relational database management system (RDBMS) for its backend performance. Web browser interface was created using HTML, CSS, Ajax, JavaScript, and jQuery, which connects MySQL terminal using several PHP scripts. A JMol visualizer (http://​www.​jmol.​org/​) and ZINC database (http://​zinc.​docking.​org/​) has been embedded in Graphical User Interface (GUI) to provide a 3D visualization and percentage similarity search against ZINC, respectively, for all PDMs. The GUI is designed to be user friendly for data query and extraction, and has been tested in all major browsers (Chrome, Firefox, Safari, and Internet Explorer) and OS platforms.

SerpentinaDB can be explored for PDMs in a number of ways through querying the database with a simple text search tool that provides various options for searching. There are three search sections available to the user with several constraints in each. Search can be performed with (i) plant part, (ii) chemical class, and (iii) physicochemical properties (Fig. 1a). Physicochemical properties search option has advanced search query options for user to select PDMs in a particular range based on MW, number of HBA, number of HBD, and number of aromatic rings. The result for given query is presented in the same page (Fig. 1b) along with information such as PDM ID, plant part, chemical name, chemical class, IUPAC names, SMILES notations, and 3D structure of PDM with associated references. Clicking the drop down arrow provides details of physicochemical and ADMET properties. Two separate links to download mol2 file and 2-Dimensional structure of PDM for given query has been provided. Also, a JMol visualizer (http://​www.​jmol.​org/​) has been embedded in GUI to provide a 3D visualization of PDM which can be further downloaded to mol2 file.

Also, each PDM from result page can be searched, to explore the neighborhood of chemical space, against ZINC database to identify analogs of natural molecules at different percentage similarity cut-off (default 90 %). In order to perform this search ZINC database, a curated collection of commercially available chemical compounds [26], is hyperlinked to result page for each PDM separately to return their structural analogs. During this similarity search natural molecules are used as scaffolds to search for similar mimetics which may have equivalent biological properties. Thus, SerpentinaDB serves as a portal to facilitate the use of natural chemical diversity for drug discovery through prospection of direct novel leads and their analogs.

SerpentinaDB provides comprehensive information of R. serpentina PDMs as a structured and integrated library. This database was developed to facilitate prospection of therapeutic molecules from this medicinally important plant. Existing repositories of natural compounds, such as NPACT [27], SuperNatural [28], Herb Ingredients’ Targets [29], and CamMedNP [30], focus on different utilitarian aspects of PDM libraries. While some of these databases emphasize on a specific disease or target-compound interactions, others cover plants of specific geography. SerpentinDB contains natural molecules of R. serpentina which is an important Himalayan medicinal plant reported for various pharmacological properties. While the reported efficacy of R. serpentina extract against hypertension has been explored very well to identify specific therapeutic PDMs, its potential against a host of other disorders (Table 1) is hitherto not pinned down to specific molecules. This database can facilitate prospection of novel leads for these disorders from the repertoire of natural molecules.

Natural molecules have been recognized to provide specific scaffolds that make them comparable to trade drugs and their potential utilization in combinatorial chemistry [4]. The MW distribution of PDMs present in SerpentinaDB has been found to follow Gaussian distribution and peaked in the range of 300-450 Da (Fig. 3a) which is similar to drug-like molecules of previously reported libraries of natural products [31]. Significant number of PDMs have HBA in the range of 3–5 with a sharp decline thereafter, as desired from drug-like molecules (Fig. 3b). Similarly, HBDs of PDMs have a peak at 2 with a sharp drop for higher values, as desired (Fig. 3b). SerpentinaDB PDMs have maximum density in the ‘Lipinski region of interest’ reflecting their drug-like properties and hence their utility in prospection of novel leads. The relevance of SerpentinaDB in drug discovery has been demonstrated with the virtual screening protocol, molecular dynamics, and ZINC similarity search for potential inhibitors of aldose reductase, a target for complications of diabetes [6]. This hypothesis driven prospection study yielded two indole alkaloids as well as their structural analogs as potential AR inhibitors (Fig. 4) [6]. This protocol serves as a demonstration of utility of SerpentinaDB for rational search of therapeutic molecules and highlights its relevance [32, 33]. Future extensions of SerpentinaDB may include 3D structure similarity search and disease associations for each PDM.