Dr. PIAS: an integrative system for assessing the druggability of protein-protein interactions
© Sugaya and Furuya; licensee BioMed Central Ltd. 2011
Received: 8 October 2010
Accepted: 9 February 2011
Published: 9 February 2011
The amount of data on protein-protein interactions (PPIs) available in public databases and in the literature has rapidly expanded in recent years. PPI data can provide useful information for researchers in pharmacology and medicine as well as those in interactome studies. There is urgent need for a novel methodology or software allowing the efficient utilization of PPI data in pharmacology and medicine.
To address this need, we have developed the 'Druggable Protein-protein Interaction Assessment System' (Dr. PIAS). Dr. PIAS has a meta-database that stores various types of information (tertiary structures, drugs/chemicals, and biological functions associated with PPIs) retrieved from public sources. By integrating this information, Dr. PIAS assesses whether a PPI is druggable as a target for small chemical ligands by using a supervised machine-learning method, support vector machine (SVM). Dr. PIAS holds not only known druggable PPIs but also all PPIs of human, mouse, rat, and human immunodeficiency virus (HIV) proteins identified to date.
The design concept of Dr. PIAS is distinct from other published PPI databases in that it focuses on selecting the PPIs most likely to make good drug targets, rather than merely collecting PPI data.
The importance of PPIs as targets for drugs, especially small molecule drugs, has increased greatly in recent years [1–4]. Over 30 PPIs have been widely studied as targets for PPI-inhibiting small ligands. Currently, a huge amount of PPI data has been rapidly accumulated in public databases and in the literature. In addition, advances in high-throughput experimental technologies have lead to a large amount of various types of omics data, which have been deposited in many databases. These PPI data and omics data require methodologies for their application to pharmacological and medicinal studies. There is an urgent need to identify novel PPIs as drug targets from the PPI data accumulated, since only about 30 druggable PPIs have been well studied to date, whereas approximately 60,000 PPIs have been identified in human. We have recently proposed integrative approaches for discovering drug target PPIs by assessing the druggability of PPIs by the use of various types of omics data [5, 6]. The application of our methods to human PPIs predicted many potentially druggable PPIs.
Several databases and web-based tools specializing in drug targets have been published. For example, TTD [7, 8], a database of known therapeutic target proteins, stores information relevant to the targets, such as tertiary structures, disease associations, pathways, and pertinent literature. PDTD , a database for in silico drug target identification, stores diverse information on drug target proteins identified by the web-based tool Target Fishing Docking. SuperPred , a web-server for drug classification, uses a similarity score between drugs/chemicals to predict drug target proteins. These drug target databases and web-servers are very useful for researchers in in silico pharmacology and medicine. All of them, however, deal only with single proteins, rather than PPIs.
Recently, two databases (2P2IDB and TIMBAL ) specializing in drug target PPIs and PPI-inhibiting chemicals have been published. 2P2IDB mainly focuses on protein/protein and protein/inhibitor interfaces in terms of various physicochemical parameters such as atom and residue properties, pocket volume, and accessible surface area . TIMBAL is a database of small molecules that inhibit protein/protein complexes, and it stores many properties of the molecules such as molecular weight, LogP value, number of rings, number of rotatable bonds, and binding affinity . 2P2IDB and TIMBAL can provide useful information to researchers developing PPI inhibitors. Both databases, however, contain only known drug target PPIs, so only a very small number of PPIs and PPI-inhibiting chemicals are stored. As a next step, in order to efficiently utilize the databases such as 2P2IDB and TIMBAL, it is needed to apply the information obtained from known drug target PPIs and their inhibitors to other PPIs not presently targeted by inhibitors.
Here we describe a novel database system, Dr. PIAS, which focuses on the druggability of PPIs. Dr. PIAS assesses the druggability of PPIs, currently not targeted by inhibitors, by utilizing the information obtained from known drug target PPIs. Dr. PIAS holds not only known drug target PPIs but also all PPIs identified to date for human, mouse, rat, and HIV proteins. In addition to information on the properties of the tertiary structures of PPI interfaces and that on the properties of drugs/chemicals related to interacting proteins, which are dealt with in 2P2IDB and TIMBAL, other properties associated with the biological function of PPIs are also included in the assessment. This is important because, to select a drug target PPI, a researcher considers not only information on the tertiary structure of the PPI and its known inhibitors but also that on the biological function of the PPI. All information on the PPIs used in the assessment is stored in Dr. PIAS. Users can search for druggable PPIs in Dr. PIAS by using various words and terms such as protein/gene name, tertiary structure, disease, pathway, and drug/chemical name as keywords.
Construction and context
Assessing the druggability of PPIs
The most distinctive characteristic of Dr. PIAS is that the system assesses the druggability of PPIs by our original SVM-based method . Thirty known drug target PPIs, including IL2/IL2RA, MDM2/TP53, and BCL2/BAK1, serve as the positive instances (Additional file 1: Table S1). These PPIs were selected from review articles focusing on druggable PPIs [1–4, 13–15]. Positive instances must satisfy both the following two criteria.
A PPI-inhibiting small chemical has been identified, and its potency as a PPI inhibitor has been validated by in vitro and/or in vivo assays.
A binding pocket for the PPI-inhibiting small chemical has been located on the tertiary structure of a protein, and it overlaps with the PPI interface.
Structural, drug/chemical, and functional attributes (Additional file 1: Table S2) of the positive instances and other PPIs in Dr. PIAS (test instances) were calculated and stored in Dr. PIAS. We used these attributes for our SVM-based method . The program package Libsvm  was used for SVM.
In previous study, we have obtained the best SVM model for discriminating the positive and negative instances, when the radial basis function kernel and the ratio of positives:negatives = 1:1 were used in machine learning by SVM . The cross validation test using the best model showed the accuracy of 80.5% (sensitivity, 81.6%; specificity, 79.4%) that was comparable to the values of accuracy in previous studies on drug target prediction . Also in Dr. PIAS, we adopt this SVM model for the assessment of the druggability of PPIs.
Data sources of PPIs
Data source of PPIs stored in Dr. PIAS.
Number of PPIs from each species stored in Dr. PIAS.
Number of PPIs
Information stored in Dr. PIAS.
Source(s) or program(s) used
Pockets on PPI interface
Accessible surface area
Amino acid composition
Amino acid sequence motifs
Chemicals associated with interacting proteins
Human diseases in OMIM
Number of interacting proteins in PPI network
GO terms and identity scores of GO terms
Gene expression profiles and similarity scores of profiles
Several properties of PPIs stored in Dr. PIAS were pre-calculated using our original and several freely available computational algorithms/programs. Among the properties of PPIs, those based on tertiary structure are the most important for in silico drug design and development. We checked whether the tertiary structure of the protein/protein complex of a PPI had been already solved. Amino acid sequence similarity searches using the computational program BLASTP  were conducted against the PDB database . If both the two interacting proteins showed sequence identities of ≥80% to distinct polypeptide chains in the same PDB entry, and the two chains physically contact each other in the tertiary structure of the protein/protein complex, the PDB entry was considered to be the tertiary structure of the PPI. Whether two chains physically contact was checked by consulting the PPI interface information in the PDBsum database .
If the tertiary structure of a PPI had been already solved, we further detected potential ligand-binding pockets that overlap with the PPI interface by using the alpha site finder implemented in the software package Molecular Operating Environment . Physicochemical and shape properties of the pockets were calculated and stored in our database (Table 3). Planarity, narrowness, and roughness of the pockets were originally defined by us [6, 26]. These properties and compactness, curvature, and amino acid composition of the pockets were calculated by using computational programs/algorithms written by us [6, 26]. Other properties were calculated by using the Molecular Operating Environment or obtained from the results of the computational program DSSP .
We retrieved information on protein domains from the Pfam database  and detected domains responsible for PPIs by consulting the iPfam database . Lists of the interacting domain pair(s) and domains of each interacting protein are stored in our database.
Recent studies have revealed that, in some groups of PPIs, disordered regions of proteins and amino acid sequence motifs in these regions are responsible for PPIs [30–33]. Among the PPIs used in the positive instances in our SVM-based method, interfaces of some PPIs such as BCL2/BAK1, BIRC4/CASP9, and MDM2/TP53 are formed by interaction between an ordered region in one protein and a disordered region in the other protein. Disordered regions in BAK1, CASP9, and TP53 in monomer changed to ordered state when protein/protein complex is formed . Some chemicals inhibiting these PPIs mimic sequence motifs in the disordered regions [34–36]. We predicted disordered regions using the computational program POODLE-L  and retrieved the information on sequence motifs from the ELM database . If the number of 'instance's of a motif in ELM was ≥2, we manually made a multiple alignment of the instances, and then created a hidden Markov model profile by using the computational program HMMER . Using the hidden Markov model profiles, the motifs were predicted for each protein by the HMMER. Lists of the motifs and disordered regions predicted by the programs are stored in our database.
In several of the known target PPIs such as ESR1/NCOA2 and GRB2/EGFR, one interacting partner (ESR1 and EGFR) is a druggable protein that has been already targeted by a drug approved by the United States Food and Drug Administration (FDA). Several other known target PPIs are novel drug targets, and both interacting partners have no FDA-approved drug targeting them. In our SVM-based method, we used the number of drugs as the PPI attributes to assess whether the fact that an interacting protein has been already targeted by existing drugs influences the selection of PPIs as drug targets. The information on the FDA-approved drugs was retrieved from the DrugBank database . The number of drugs targeting each interacting partner protein of a PPI was counted and stored in our database together with lists of drugs.
The information on chemicals that experimentally assayed for the activity to each of the two interacting proteins was retrieved from the ChEMBL database .
Information on human diseases caused when a protein is heritably or somatically mutated is essential for assessing the druggability of the protein. We retrieved information on human diseases from the OMIM database . When using the information in our SVM-based method, it was transformed to a score of 0 or 1 . The information is scored as 1 if both two interacting proteins of a PPI are implicated in OMIM-registered diseases (not limited to the same disease). The attribute is scored as 0 if only one interacting protein is implicated in a disease or if neither interacting protein is implicated in diseases. Lists of diseases associated with each interacting protein are stored in our database.
To repress a disease state with drugs, it would be desirable in some cases to target the proteins that function as 'hubs' in a PPI network. In other cases, targeting proteins that function in a peripheral part of a network could be more feasible for the treatment of a disease. After the PPI network was constructed based on the PPI data stored in Dr. PIAS, the number of all interacting proteins for each partner of a PPI was counted and stored in our database.
As with the number of interacting proteins, proteins that function in a large number of biological pathways may be more desirable targets for therapeutic intervention for some diseases, while proteins involved in a limited number of pathways may be more desirable targets for other diseases. We retrieved information on biological pathways from the KEGG  and PID  databases. We counted the number of pathways in which a protein is involved and stored these numbers in our database together with lists of pathways.
To assess the degree of similarity in biological function of the two interacting proteins, we utilized Gene Ontology (GO) . GO terms assigned to proteins in Dr. PIAS were retrieved from the GO database . Based on the GO terms, we calculated the identity scores of GO terms between the two interacting proteins according to equation S3 in Additional file four in . The identity scores and lists of GO terms assigned to each protein are stored in our database.
For a protein to be selected as a drug target, it is advantageous if the protein functions in a limited number of tissues/organs including the tissues/organs that develop the disease. Because of a scarcity of information on protein expression profiles in human, we utilized gene expression profiles. Information on gene expression profiles was retrieved from the UniGene database . To assess the degree of similarity between the gene expression profiles of the two interacting proteins in a PPI, we calculated the similarity scores of expression profiles between the two genes according to equation S4 in Additional file 4 in . The similarity scores and gene expression profiles are stored in our database.
The number of paralogs is an important factor in order for a protein to be selected as a drug target, since the researchers must consider potential adverse effects caused by the drug binding to non-target paralogs. Target proteins with a large number of paralogs may be associated with more severe adverse effects. The information on paralogs was retrieved from the KEGG and PIRSF  databases. We stored the number of paralogs of each protein and lists of paralogs in our database.
All information described above is stored in a relational database (MySQL).
Figure 2C shows a main search result summarizing PPIs and their druggability scores. The columns of the druggability score are colored depending on the score (≥0.99, red; <0.99 and ≥0.9, magenta; <0.9 and ≥0.8, hotpink; <0.8 and ≥0.7, pink). The threshold of the coloring of ≥0.9 (red and magenta) is set based on the average value of the druggability scores of the positive instances (known drug target PPIs) used in the druggability assessment (see Figure 5 in ). Other thresholds are set arbitrarily. The assessment of the druggability of PPIs is conducted in four ways: three of the four use only the structural, drug/chemical, or functional attributes of the PPIs, respectively, while the fourth uses all attributes in our SVM-based method (Additional file 1: Table S2). These four ways correspond to the four columns of the druggability score in Figure 2C.
If a user selects the 'Draw PPI network' radio button in the 'Advanced search' form, a PPI network derived from the search result is drawn by the computational program eXpanda  (Figure 2D). Nodes and edges of the network represent proteins and PPIs, respectively. The PPI network reflects information stored in Dr. PIAS. If the tertiary structure of the protein/protein complex of a PPI has been solved, the edge representing the PPI is shown as a thick line. Edges of the network are colored according to the druggability score (calculated using all attributes). The size of the nodes is approximately proportional to the number of interacting partner proteins in Dr. PIAS. Nodes circled with a thick line represent proteins whose tertiary structures (but not necessarily those of the protein/protein complex) have been already solved. Nodes colored lightbrown represent proteins that are registered as disease-implicated in OMIM. Nodes shown by a diamond shape indicate proteins registered as known drug targets in DrugBank.
By clicking the headers of the summary table shown in Figure 2C, users can see details of the information stored in Dr. PIAS. Figure 2E is an example of the level of detail regarding a potential ligand-binding pocket on a PPI interface, which is used for assessing the druggability of a PPI. The pocket is shown by atoms colored red. The image was generated by using the software Jmol .
In this section, the usage of Dr. PIAS is illustrated by applying it to search for potentially druggable PPIs from two points of view: one is from a disease point of view and the other is from a protein.
Searching for druggable PPIs implicated in lung cancer
Lung cancer is one of the most common cancers in both men and women worldwide . Many studies to develop drugs for the treatment of lung cancer have been conducted both in the pharmaceutical industry and in academia, but only a few drugs have been approved by the FDA [51, 52]. The identification of novel drug targets for lung caner would prompt the development of drugs targeting this cancer.
A PPI is derived from human.
The PPI is involved in biological pathways associated with lung caner.
Genes coding interacting proteins are expressed in lung cancer cells.
Tertiary structure of the protein complex of the PPI has been solved.
List of PPIs assessed as potentially druggable for lung cancer.
Interestingly, the list of PPIs in Table 4 includes EGFR/TGFA, and the PPI has a high druggability score of 0.9393. TGFA is a member of the EGF family, and has a tertiary structure similar to EGF. TGFA forms a protein/protein complex with EGFR in a manner similar to EGFR/EGF complex . EGFR/EGF PPI is the target for the FDA-approved anti-EGFR antibody, cetuximab, for the treatment of several cancers such as colorectal cancer and head and neck cancer [51, 52]. EGFR/EGF is assessed as highly druggable by our method in three of the four assessments. A druggability score of 0.8712 is obtained when only structural attributes are used, whereas a score of 0.9997 is obtained when functional attributes are used, and a score of 0.8724 when all attributes are used (Additional file 1: Table S3). Thus, EGFR/TGFA may be a drug target for cancers including lung cancer. In Table S3, another PPI, ERBB2(HER2)/ERBB2(HER2), which has been already targeted by a FDA-approved drug, is included. The anti-HER2 antibody, trastuzumab, has been developed for the therapeutic intervention of breast cancer [51, 52]. Although these FDA-approved drugs, cetuximab and trastuzumab, are not small molecule drugs, these results indicate that Dr. PIAS is effective in predicting druggable PPIs. PPIs in Table 4 other than GRB2/VAV1 and EGFR/TGFA may be potential drug targets for lung cancer in future studies.
Searching for druggable PPIs associated with BCL-XL
As described in the former section, 2P2IDB and TIMBAL have been already published that hold drug target PPIs and their inhibitors. By using an identical protein as a query when searching for druggable PPIs in Dr. PIAS and the two databases, we demonstrate the differences and similarities in output results obtained from the three databases.
BCL-XL protein was adopted as a query. The PPIs of BCL-XL with BAX, BAK, and BID proteins are well-studied drug target PPIs , and both 2P2IDB and TIMBAL contain the PPIs. From 2P2IDB, a user can obtain the information on 8 tertiary structures (as of 2010/12) of BCL-XL/BAK protein/protein and BCL-XL/inhibitor complexes and the information on various physicochemical properties of the inhibitor-binding pockets. If a user search for BCL-XL in TIMBAL, the user can obtain the information on 26 chemicals (as of 2010/12) inhibiting the PPIs of BCL-XL with BAX, BAK, and BID. The two databases provide users with the structural and drug/chemical information on already-studied drug target PPIs.
When a user use the protein name 'BCL2L1' (a synonym of BCL-XL) as a keyword in searching Dr. PIAS, the user can obtain 66 PPIs (as of 2010/12) of BCL-XL with many other proteins as well as with BAX, BAK, and BID (Additional file 1: Table S4). Distinct from 2P2IDB and TIMBAL, Dr. PIAS assesses the druggability of all PPIs associated with BCL-XL as well as already-studied drug target PPIs and provides structural, drug/chemical, and functional information on all PPIs. Among the 66 PPIs, 5 PPIs (BAD/BCL2L1, BAK1/BCL2L1, BCL2L1/BCL2L1, BCL2L1/BCL2L11, and BCL2L1/Bcl2l11) have the tertiary structures of protein/protein complexes in the PDB, and the druggability assessments were conducted in all four ways (Additional file 1: Table S4). BAK1/BCL2L1 has the highest druggability score of 0.7674 when all attributes were used for the assessment. Like 2P2IDB, the information on the physicochemical and shape properties of the ligand-binding pocket on BAK1/BCL2L1 PPI interface can be obtained from Dr. PIAS (Table 3). Like TIMBAL, a user can obtain the drug/chemical information from Dr. PIAS. Unlike 2P2IDB and TIMBAL, Dr. PIAS provide a user with the information on biological function of BAK1 and BCL2L1 proteins (Table 3). BCL2L1 is involved in apoptosis pathway, focal adhesion pathway, and some cancer-related pathways (based on KEGG). BAK1 functions as a direct p53 effector (based on PID). The gene encoding BCL2L1 is expressed in many body sites (34/44 in UniGene). The gene encoding BAK1 is also the case (32/44 in UniGene). In PIRSF, BCL2L1 has 10 paralogs, while BAK1 has 5 paralogs. In addition to the structural and drug/chemical information, the functional information described above can be also helpful for a researcher to select drug target PPIs.
Comparisons with existing databases
Currently, there exist two databases (2P2IDB and TIMBAL) that focus on drug target PPIs. These databases hold known drug target PPIs and their inhibitors. In contrast, Dr. PIAS holds all PPIs of human, mouse, rat, and HIV proteins identified to date as well as known drug target PPIs. For each PPI, we assessed the druggability by the SVM-based method by using structural, drug/chemical, and functional attributes of the PPIs. These two characteristics of Dr. PIAS are what make Dr. PIAS distinct from other existing databases. 2P2IDB and TIMBAL primarily focus on the structural aspects of protein pockets and the chemical properties of PPI-inhibiting ligands, respectively. The information on the biological function of a protein is also essential for selecting drug target PPIs. Dr. PIAS stores functional information on interacting proteins, such as disease associations, pathways, GO terms, gene expression profiles, and paralogs. Therefore, Dr. PIAS can help researchers select drug target PPIs by evaluating each PPI from the three viewpoints of the tertiary structures of protein/protein complexes, drugs/chemicals relevant to interacting proteins, and the biological roles of PPIs in living cells. A cross reference of the three databases can provide researchers with a synergistic power to prompt studies on drug target PPIs and chemicals inhibiting them.
Many databases of PPIs have been published, all with the purpose of collecting as many PPIs as possible from the literature and already-published similar databases. The main purpose for developing Dr. PIAS was not to merely collect PPI data but to select useful PPIs (in this case, as potential drug targets) from the collected data. In this sense, the design concept of Dr. PIAS is completely different from that of all other PPI databases published to date.
The number of PPIs for which small molecule inhibitors have been discovered has gradually increased over the last decade. This trend will continue. As more information on known drug target PPIs accumulates, our assessment system based on information obtained from known target PPIs will be improved. We will intensively incorporate information on novel drug target PPIs to make Dr. PIAS more useful for researchers focusing on the development of PPI-inhibiting drugs. In addition, PPI data and all omics information will be updated every half year. The number of PPIs stored in Dr. PIAS will rapidly increase due to the accumulation of PPI data in public databases and in the literature.
Dr. PIAS is a database system aimed at assessing the druggability of PPIs. Of the huge number of currently unidentified PPIs, there could be many latent PPIs that are highly druggable. Dr. PIAS will aid the efficient discovery of these druggable PPIs from the continuously growing amount of PPI data.
Availability and requirements
Dr. PIAS is available at http://asp.gridasp.net/drpias/index.php. Academic non-profit users can freely access all of the contents stored in Dr. PIAS without paying a licensing fee. Commercial and for-profit users must obtain a license to access Dr. PIAS by paying a licensing fee to and entering into a license agreement with Beyond Computing, Co. Ltd. and PharmaDesign, Inc.
List of abbreviations
- Dr. PIAS:
druggable protein-protein interaction assessment system
United States Food and Drug Administration
human immunodeficiency virus
support vector machine.
We gratefully acknowledge the staff of Beyond Computing, Co. Ltd. for providing us with a computational server to execute Dr. PIAS with high computational performance.
- Arkin MR, Wells JA: Small-molecule inhibitors of protein-protein interactions: progressing towards the dream. Nat Rev Drug Discov 2004, 3: 301–317. 10.1038/nrd1343View ArticlePubMedGoogle Scholar
- Pagliaro L, Felding J, Audouze K, Nielsen SJ, Terry RB, Krog-Jensen C, Butcher S: Emerging classes of protein-protein interaction inhibitors and new tools for their development. Curr Opin Chem Biol 2004, 8: 442–449. 10.1016/j.cbpa.2004.06.006View ArticlePubMedGoogle Scholar
- Zhao L, Chmielewski J: Inhibiting protein-protein interactions using designed molecules. Curr Opin Struct Biol 2005, 15: 31–34. 10.1016/j.sbi.2005.01.005View ArticlePubMedGoogle Scholar
- Wells JA, McClendon CL: Reaching for high-hanging fruit in drug discovery at protein-protein interfaces. Nature 2007, 450: 1001–1009. 10.1038/nature06526View ArticlePubMedGoogle Scholar
- Sugaya N, Ikeda K, Tashiro T, Takeda S, Otomo J, Ishida Y, Shiratori A, Toyoda A, Noguchi H, Takeda T, Kuhara S, Sakaki Y, Iwayanagi T: An integrative in silico approach for discovering candidates for drug-targetable protein-protein interactions in interactome data. BMC Pharmacol 2007, 7: 10. 10.1186/1471-2210-7-10PubMed CentralView ArticlePubMedGoogle Scholar
- Sugaya N, Ikeda K: Assessing the druggability of protein-protein interactions by a supervised machine-learning method. BMC Bioinformatics 2009, 10: 253. 10.1186/1471-2105-10-263View ArticleGoogle Scholar
- Chen X, Ji ZL, Chen YZ: TTD: Therapeutic Target Database. Nucleic Acids Res 2002, 30: 412–415. 10.1093/nar/30.1.412PubMed CentralView ArticlePubMedGoogle Scholar
- Zhu F, Han B, Kumar P, Liu X, Ma X, Wei X, Huang L, Guo Y, Han L, Zheng C, Chen Y: Update of TTD: Therapeutic Target Database. Nucleic Acids Res 2010, 38: D787–791. 10.1093/nar/gkp1014PubMed CentralView ArticlePubMedGoogle Scholar
- Gao Z, Li H, Zhang H, Liu X, Kang L, Luo X, Zhu W, Chen K, Wang X, Jiang H: PDTD: a web-accessible protein database for drug target identification. BMC Bioinformatics 2008, 9: 104. 10.1186/1471-2105-9-104PubMed CentralView ArticlePubMedGoogle Scholar
- Dunkel M, Günther S, Ahmed J, Wittig B, Preissner R: SuperPred: drug classification and target prediction. Nucleic Acids Res 2008, 36: W55-W59. 10.1093/nar/gkn307PubMed CentralView ArticlePubMedGoogle Scholar
- Bourgeas R, Basse MJ, Morelli X, Roche P: Atomic analysis of protein-protein interfaces with known inhibitors: the 2P2I database. PLoS ONE 2010, 5: e9598. 10.1371/journal.pone.0009598PubMed CentralView ArticlePubMedGoogle Scholar
- Higueruelo AP, Schreyer A, Bickerton GR, Pitt WR, Groom CR, Blundell TL: Atomic interactions and profile of small molecules disrupting protein-protein interfaces: the TIMBAL database. Chem Biol Drug Des 2009, 74: 457–467. 10.1111/j.1747-0285.2009.00889.xView ArticlePubMedGoogle Scholar
- Toogood PL: Inhibition of protein-protein association by small molecules: approaches and progress. J Med Chem 2007, 45: 1543–1558. 10.1021/jm010468sView ArticleGoogle Scholar
- Loregian A, Palù G: Disruption of protein-protein interactions: towards new targets for chemotherapy. J Cell Physiol 2005, 204: 750–762. 10.1002/jcp.20356View ArticlePubMedGoogle Scholar
- Fletcher S, Hamilton AD: Protein-protein interaction inhibitors: small molecules from screening techniques. Curr Top Med Chem 2007, 7: 922–927. 10.2174/156802607780906735View ArticlePubMedGoogle Scholar
- LIBSVM-A Library for Support Vector Machines-[http://www.csie.ntu.edu.tw/~cjlin/libsvm/]
- Maglott D, Ostell J, Pruitt KD, Tatusova T: Entrez Gene: gene-centered information at NCBI. Nucleic Acids Res 2007, 35: D26-D31. 10.1093/nar/gkl993PubMed CentralView ArticlePubMedGoogle Scholar
- Genome Network Platform[http://genomenetwork.nig.ac.jp/download/dataset_e.html]
- Stelzl U, Worm U, Lalowski M, Haenig C, Brembeck FH, Goehler H, Stroedicke M, Zenkner M, Schoenherr A, Koeppen S, Timm J, Mintzlaff S, Abraham C, Bock N, Kietzmann S, Goedde A, Toksöz E, Droege A, Krobitsch S, Korn B, Birchmeier W, Lehrach H, Wanker EE: A human protein-protein interaction network: a resource for annotating the proteome. Cell 2005, 122: 957–968. 10.1016/j.cell.2005.08.029View ArticlePubMedGoogle Scholar
- Lim J, Hao T, Shaw C, Patel AJ, Szabó G, Rual JF, Fisk CJ, Li N, Smolyar A, Hill DE, Barabási AL, Vidal M, Zoghbi HY: A protein-protein interaction network for human inherited ataxias and disorders of Purkinje cell degeneration. Cell 2006, 125: 801–814. 10.1016/j.cell.2006.03.032View ArticlePubMedGoogle Scholar
- Ramani AK, Li Z, Hart GT, Carlson MW, Boutz DR, Marcotte EM: A map of human protein interactions derived from co-expression of human mRNAs and their orthologs. Mol Syst Biol 2008, 4: 180. 10.1038/msb.2008.19PubMed CentralView ArticlePubMedGoogle Scholar
- Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res 1997, 25: 3389–3402. 10.1093/nar/25.17.3389PubMed CentralView ArticlePubMedGoogle Scholar
- Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE: The Protein Data Bank. Nucleic Acids Res 2000, 28: 235–242. 10.1093/nar/28.1.235PubMed CentralView ArticlePubMedGoogle Scholar
- Chemical Computing Group[http://www.chemcomp.com/]
- Kabsch W, Sander C: Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers 1983, 22: 2577–2637. 10.1002/bip.360221211View ArticlePubMedGoogle Scholar
- Dr. PIAS Help[http://asp.gridasp.net/drpias/pias_help.php]
- Laskowski RA, Chistyakov VV, Thornton JM: PDBsum more: new summaries and analyses of the known 3 D structures of proteins and nucleic acids. Nucleic Acids Res 2005, 33: D266–268. 10.1093/nar/gki001PubMed CentralView ArticlePubMedGoogle Scholar
- Finn RD, Mistry J, Tate J, Coggill P, Heger A, Pollington JE, Gavin OL, Gunasekaran P, Ceric G, Forslund K, Holm L, Sonnhammer EL, Eddy SR, Bateman A: The Pfam protein families database. Nucleic Acids Res 2010, 38: D211-D222. 10.1093/nar/gkp985PubMed CentralView ArticlePubMedGoogle Scholar
- Finn RD, Marshall M, Bateman A: iPfam: visualization of protein-protein interactions in PDB at domain and amino acid resolutions. Bioinformatics 2005, 21: 410–412. 10.1093/bioinformatics/bti011View ArticlePubMedGoogle Scholar
- Neduva V, Linding R, Su-Angrand I, Stark A, de Masi F, Gibson TJ, Lewis J, Serrano L, Russell RB: Systematic discovery of new recognition peptides mediating protein interaction networks. PLoS Biol 2005, 3: e405. 10.1371/journal.pbio.0030405PubMed CentralView ArticlePubMedGoogle Scholar
- Hegyi H, Schad E, Tompa P: Structural disorder promotes assembly of protein complexes. BMC Struct Biol 2007, 7: 65. 10.1186/1472-6807-7-65PubMed CentralView ArticlePubMedGoogle Scholar
- Fuxreiter M, Tompa P, Simon I: Local structural disorder imparts plasticity on linear motifs. Bioinformatics 2007, 23: 950–956. 10.1093/bioinformatics/btm035View ArticlePubMedGoogle Scholar
- Kim PM, Sboner A, Xia Y, Gerstein M: The role of disorder in interaction networks: a structural analysis. Mol Sys Biol 2008, 4: 179.Google Scholar
- Cheng Y, LeGall T, Oldfield CJ, Mueller JP, Van YY, Romero P, Cortese MS, Uversky VN, Dunker AK: Rational drug design via intrinsically disordered protein. Trends Biotechnol 2006, 24: 435–442. 10.1016/j.tibtech.2006.07.005View ArticlePubMedGoogle Scholar
- Lessene G, Czabotar PE, Colman PM: BCL-2 family antagonists for cancer therapy. Nat Rev Drug Discov 2008, 7: 989–1000. 10.1038/nrd2658View ArticlePubMedGoogle Scholar
- Sun H, Nikolovska-Coleska Z, Yang CY, Xu L, Tomita Y, Krajewski K, Roller PP, Wang S: Structure-based design, synthesis, and evaluation of conformationally constrained mimetics of the second mitochondria-derived activator of caspase that target the X-linked inhibitor of apoptosis protein/caspase-9 interaction site. J Med Chem 2004, 47: 4147–4150. 10.1021/jm0499108View ArticlePubMedGoogle Scholar
- Hirose S, Shimizu K, Kanai S, Kuroda Y, Noguchi T: POODLE-L: a two-level SVM prediction system for reliably predicting long disordered regions. Bioinformatics 2007, 23: 2046–2053. 10.1093/bioinformatics/btm302View ArticlePubMedGoogle Scholar
- Gould CM, Diella F, Via A, Puntervoll P, Gemünd C, Chabanis-Davidson S, Michael S, Sayadi A, Bryne JC, Chica C, Seiler M, Davey NE, Haslam N, Weatheritt RJ, Budd A, Hughes T, Pas J, Rychlewski L, Travé G, Aasland R, Helmer-Citterich M, Linding R, Gibson TJ: ELM: the status of the 2010 eukaryotic linear motif resource. Nuclic Acids Res 2010, 38: D167-D180. 10.1093/nar/gkp1016View ArticleGoogle Scholar
- HMMER: biosequence analysis using profile hidden Markov models.[http://hmmer.janelia.org/]
- Wishart DS, Knox C, Guo AC, Cheng D, Shrivastava S, Tzur D, Gautam B, Hassanali M: DrugBank: a knowledgebase for drugs, drug actions and drug targets. Nucleic Acids Res 2008, 36: D901-D906. 10.1093/nar/gkm958PubMed CentralView ArticlePubMedGoogle Scholar
- Hamosh A, Scott AF, Amberger JS, Bocchini CA, McKusick VA: Online Mendelian Inheritance in Man (OMIM), a knowledgebase of human genes and genetic disorders. Nucleic Acids Res 2005, 33: D514-D517. 10.1093/nar/gki033PubMed CentralView ArticlePubMedGoogle Scholar
- Kanehisa M, Goto S, Furumichi M, Tanabe M, Hirakawa M: KEGG for representation and analysis of molecular networks involving diseases and drugs. Nucleic Acids Res 2010, 38: D355-D360. 10.1093/nar/gkp896PubMed CentralView ArticlePubMedGoogle Scholar
- Schaefer CF, Anthony K, Krupa S, Buchoff J, Day M, Hannay T, Buetow KH: PID: The Pathway Interaction Database. Nucleic Acids Res 2009, 37: D674-D679. 10.1093/nar/gkn653PubMed CentralView ArticlePubMedGoogle Scholar
- The Gene Ontology Consortium: The Gene Ontology in 2010: extentions and refinements. Nucleic Acids Res 2010, 38: D331-D335. 10.1093/nar/gkp1018PubMed CentralView ArticleGoogle Scholar
- Pontius JU, Wagner L, Schuler GD: UniGene: a unified view of the transcriptome. In The NCBI Handbook. Bethesda(MD): National Center for Biotechnology Information; 2003.Google Scholar
- Wu CH, Nikolskaya A, Huang H, Yeh LS, Natale DA, Vinayaka CR, Hu ZZ, Mazumder R, Kumar S, Kourtesis P, Ledley RS, Suzek BE, Arminski L, Chen Y, Zhang J, Cardenas JL, Chung S, Castro-Alvear J, Dinkov G, Barker WC: PIRSF: family classification system at the Protein Information Resource. Nucleic Acids Res 2004, 32: D112-D114. 10.1093/nar/gkh097PubMed CentralView ArticlePubMedGoogle Scholar
- Negishi Y, Nakamura H, Yachie N, Saito R, Tomita M: eXpanda: an integrated platform for network analysis and visualization. In Silico Biol 2007, 7: 141–144.PubMedGoogle Scholar
- Jmol: an open-source Java viewer for chemical structures in 3 D[http://jmol.sourceforge.net/]
- Jemal A, Siegel R, Ward E, Hao Y, Xu J, Murray T, Thun MJ: Cancer statistics, 2008. CA Cancer J Clin 2008, 58: 71–96. 10.3322/CA.2007.0010View ArticlePubMedGoogle Scholar
- Lurje G, Lenz H-J: EGFR signaling and drug discovery. Oncology 2009, 77: 400–410. 10.1159/000279388View ArticlePubMedGoogle Scholar
- Katzel JA, Fanucchi MP, Li Z: Recent advances of novel targeted therapy in non-small cell lung cancer. J Hematol Oncol 2009, 2: 2. 10.1186/1756-8722-2-2PubMed CentralView ArticlePubMedGoogle Scholar
- Bogan AA, Thorn KS: Anatomy of hot spots in protein interfaces. J Mol Biol 1998, 280: 1–9. 10.1006/jmbi.1998.1843View ArticlePubMedGoogle Scholar
- Brinda KV, Kannan N, Vishveshwara S: Analysis of homodimeric protein interfaces by graph-spectral methods. Protein Eng 2002, 15: 265–277. 10.1093/protein/15.4.265View ArticlePubMedGoogle Scholar
- Dharmawardana PG, Peruzzi B, Giubellino A, Burke TR Jr, Bottaro DP: Molecular targeting of growth factor receptor-bound 2 (Grb2) as an anti-cancer strategy. Anticancer Drugs 2006, 17: 13–20. 10.1097/01.cad.0000185180.72604.acView ArticlePubMedGoogle Scholar
- Garrett TP, McKern NM, Lou M, Elleman TC, Adams TE, Lovrecz GO, Zhu HJ, Walker F, Frenkel MJ, Hoyne PA, Jorissen RN, Nice EC, Burgess AW, Ward CW: Crystal structure of a truncated epidermal growth factor receptor extracellular domain bound to transforming growth factor alpha. Cell 2002, 110: 763–773. 10.1016/S0092-8674(02)00940-6View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.