PDTD: a web-accessible protein database for drug target identification

Until 2000, only ~500 drug targets had been reported [1], among which only 120 drug targets are actually marketed [2]. The completion of human genome and numerous pathogen genomes suggests that there are 30,000 to 40,000 genes and at least the same number of proteins, and many of these proteins are potential targets for drug discovery. It has been estimated that there are more than 2,000 potential drug targets with at least one drug candidate in clinical trial [2, 3]. This is a reservoir for drug discovery and target identification. However, how to extensively utilize this source is a challenge. Expressing all these proteins and screening compounds against the corresponding models constructed based on the proteins is extremely unpractical, because it is intolerably expensive and time-consuming. Recent promising advancement in docking-based virtual screening has demonstrated the efficiency of this approach in discovering lead (active) compounds [4, 5]. On the other hand, reverse (or inverse) docking approaches have become promising computational tools to find the probable target proteins for active compounds, natural products or old drugs [6–10]. Both these two researches need the information of target proteins, in particular the information of structures and active sites. However, such information of most drug targets is dispersedly deposited in the literatures or other databases like Protein Data Bank (PDB). Therefore, it is in dire need of a database containing comprehensive information of the potential target proteins.

Recently, some notable efforts have been made to partially satisfy this requirement. The Therapeutic Target Database (TTD) is one such example [11], which provides information about the known therapeutic targets, disease conditions and the corresponding drugs. DrugBank is a bioinformatics/cheminformatics resource that combines detailed drug data with comprehensive drug target information [12]. A number of ligand-protein interaction databases have also emerged including LigBase [13], PDBsite [14], SitesBase [15], MSDsite [16], PDB-Ligand [17] and AffinDB [18]. Unfortunately, these databases were not specifically designed for discovering new leads by using virtual screening approaches and new targets by using reverse docking. They also cannot be used to figure out specific pharmaceutical information related to the secondary effects of an old drug (i.e. new pharmacological usage) or an existing target (i.e. new pharmacological or toxic relevance). Ideally, a target database may provide not only abundant information about the potential target proteins such as 3D structures, binding (active) sites, biological (pharmacological) functions, related diseases, but also appropriate computational tools to mine the information about targets. Herein, we present a web-accessible protein database, PDTD (P otential D rug T arget D atabase). Integrated with our reverse docking server, TarFisDock [8], PDTD is a valuable platform for target identification.

Fundamentally, PDTD has dual functions of querying drug target information and identifying the potential binding proteins of an active compound or an existing drug by using reverse docking approach. Accordingly, PDTD contains two sub-databases types, one is the structural sub-database and another is the informatics sub-database. All data are associated with a relational database implemented using MySQL and can be queried through web interface. Through three computational engines, search engine, visualization engine and TarFisDock, users can implement interactive query and computation with the PDTD (Figure 1). The structural sub-database stores each protein in both PDB format and mol2 format with Amber charges; sequence and active site information were also included in the structural sub-database. The informatics database stores the data of target categories, related disease information, biological functions and associated regulating (signaling) pathways. PDTD currently contains >1100 entries covering the information of >800 known drug targets.

The target proteins in PDTD were selected from scientific literatures [1, 19–21] and several online databases such as TTD [11], DrugBank [12] and Thomson Pharma [22]. Since PDTD is designed to search the probable binding proteins for new active compounds or existing drugs by using reverse docking, it only contains the proteins with known 3D structures determined experimentally by the X-ray crystallographic or NMR methods. The coordinates of proteins were isolated from the PDB. Since not all PDB structures are of equal quality, a protein structure is selected according to the following criteria when it has several redundant records in PDB: (i) select the structure without mutation and missing residues around the active site; (ii) select the structure with high resolution; (iii) select the structure complexed with ligand. For each selected protein in PDTD, amino acid residues within 6.5 Å around the bound ligand were used to define the binding (active) site. A PDB entry could contain data on a number of binding sites. If so, a separate entry was generated in the PDTD to accommodate each of the sites. HETATM records in PDB files were used to define the ligands. PDB and mol2 files of each protein were also stored in the structural sub-database. All kinds of structures for a drug target can be visualized using the "Jmol" JAVA applet [23].

Most of drug targets in PDTD have been collected with single structure (709 cases). Since our reverse docking program, TarFisDock, has not considered the flexibility of proteins, PDTD contains some redundancy for the flexible proteins. For example, HIV-1 protease has 27 entries and dihydrofolate reductase (DHFR) has 14 entries in the database. The redundancy of each target is shown in Figure 2a.

According to therapeutic areas, the drug targets may be categorized into 14 types (Figure 2b). It is convenient for users to custom a special list when they predict potential binding targets for small molecules using TarFisDock. Among targets having explicit therapeutic functions, the targets related neoplastic disease are most populated, following are hormones and hormones antagonists related targets. Targets related to viral infections are also major contributors in PDTD. The distribution of biochemical classification is shown in Figure 2c, indicating that PDTD mainly consists of enzymes, receptors, ion channels, transporters, nuclear receptors, binding protein, structural proteins, signaling proteins, factors, regulators and hormones. The targets which can not be assigned into any of these biochemical classes are grouped into an "unknown" class. The selected drug targets are enriched in enzymes (80.2%). G-protein coupled receptors (GPCR) and other receptors which account for most drug targets seldom have crystal structures, resulting that the ratio of receptor targets in PDTD is only 4.2%.

In summary, PDTD is a comprehensive, web-accessible database of drug targets, which focuses on those drug targets with known 3D-structures. By far, PDTD has collected >1100 entries covering >800 known and potential drug targets from the Protein Data Bank. PDB structure, mol2 file and active site information of each drug target were extracted from the crystal structure, and all the information can be viewed with molecular visualization tools and can be downloaded freely by users. Drug targets of PDTD were categorized by two criteria: therapeutic areas and biochemical criteria. Each target was carefully annotated by browsing several databases, such as DrugBank, TTD, and UniProt. All these information were stored in informatics sub-database, which was associated to structural sub-database with a relational database. Users can also use our reverse docking program to search PDTD for finding the possible binding protein(s) of a small molecule.

One drug molecule may interact with several targets including targets associated with side effect (toxicity). By searching PDTD, TarFisDock may provide multiple options of the binding proteins for a small molecule. These clues are useful for further experimental test in discovering new targets and new pharmacological efficacy or toxicity for an existing drug. Thus, combining with TarFisDock, PDTD is a good web-accessible protein database for identifying drug targets and for discovering new usages of old drugs [27, 28]. The user just inputs the structure of querying compound and customizes a target list from PDTD (a list of all the targets is recommended), TarFisDock may provide possible binding proteins of the compound. The results can be downloaded in the form of mol2 file with the binding pose of each compound and a list of potential binding targets according to their ranking scores.

PDTD will be updated continuously. We intend to classify the drug targets according more completely to their biological functions, which will be achieved by integrating and/or linking PDTD with other bioinformatics databases. For example, links can be directed to the databases of SOURCE [29] and Gene Ontology [30] for more descriptions of functional annotations, ontologies, and gene expression data.

PDTD is freely available for academic user at http://www.dddc.ac.cn/pdtd. To download the files of PDTD, users must complete a simple registration process and agree not to republish the data without explicit permission. Users are invited to contact us through the 'Contact' link and to participate in the user forum at http://www.dddc.ac.cn/tarfisdock/forum/.