Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

PLIP: fully automated proteinligand interaction profiler

PLIP: fully automated proteinligand interaction profiler Published online 14 April 2015 Nucleic Acids Research, 2015, Vol. 43, Web Server issue W443–W447 doi: 10.1093/nar/gkv315 PLIP: fully automated protein–ligand interaction profiler 1 1 1 1,2 Sebastian Salentin , Sven Schreiber , V. Joachim Haupt , Melissa F. Adasme and 1,* Michael Schroeder 1 2 Biotechnology Center (BIOTEC), TU Dresden, Tatzberg 47-49, 01307 Dresden, Germany and Escuela de Ingenier´ ıa en Bioinformatica, ´ Universidad de Talca, Avda. Lircay s/n Talca, 3460000, Chile Received January 29, 2015; Revised March 17, 2015; Accepted March 28, 2015 ABSTRACT The characterization of interactions in protein–ligand complexes is essential for research in structural bioinformatics, drug discovery and biology. How- ever, comprehensive tools are not freely avail- able to the research community. Here, we present the protein–ligand interaction profiler (PLIP), a novel web service for fully automated detection and visualization of relevant non-covalent protein– ligand contacts in 3D structures, freely available at projects.biotec.tu-dresden.de/plip-web. The input is either a Protein Data Bank structure, a protein or ligand name, or a custom protein–ligand com- plex (e.g. from docking). In contrast to other tools, Figure 1. Example of interaction diagram generated with PLIP: Vari- the rule-based PLIP algorithm does not require any cella zoster virus thymidine kinase (1OSN) binding the antiherpes drug structure preparation. It returns a list of detected in- brivudine-monophosphate. The binding is dominated by a double - stacking and polar interactions at the terminal regions of the ligand. teractions on single atom level, covering seven in- teraction types (hydrogen bonds, hydrophobic con- tacts, pi-stacking, pi-cation interactions, salt bridges, lar recognition and protein function or to develop and op- water bridges and halogen bonds). PLIP stands out timize lead compounds. On the other hand, comparative by offering publication-ready images, PyMOL ses- high-throughput analyses of interaction patterns can con- sion files to generate custom images and parsable siderably improve protein–ligand docking or virtual screen- result files to facilitate successive data processing. ing (4) and thus enhance in silico approaches in drug discov- The full python source code is available for download ery. However, the scientific community lacks freely available on the website. PLIP’s command-line mode allows for tools to detect frequent non-covalent protein–ligand inter- high-throughput interaction profiling. actions such as hydrophobic interactions, hydrogen bonds, salt bridges and -stacking (5). To this end, we herein present the free and open-source protein–ligand interac- INTRODUCTION tion profiler (PLIP), a fully automated and easy to use web The Protein Data Bank (PDB) (1) hosts nearly 100 000 de- server and command-line tool for protein–ligand interac- posited protein structures, with over 75% of them solved tion detection (Figure 1). in complex with a small molecule ligand. Binding of a PLIP is complementary to other state-of-the-art web ligand to its host protein requires a specific arrangement tools such as SwissDock (6), GalaxySite (7)orProBiS(8) of attractive, typically non-covalent contacts between both and can thus be applied in evaluation of docking results molecules. With such rich data at hand, we can gain deep (Figure 4), drug design (Figure 5), binding site similarity insights into how ligands interact with their protein targets assessment (3,9) and drug repositioning (10). The PLIP (2,3). Detailed characterization of these interaction pat- web service allows for comprehensive detection and visu- terns in individual cases is crucial to understand molecu- alization of protein–ligand interaction patterns from 3D To whom correspondence should be addressed. Tel: +49 351 46340060; Fax: +49 351 46340061; Email: ms@biotec.tu-dresden.de C The Author(s) 2015. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. W444 Nucleic Acids Research, 2015, Vol. 43, Web Server issue be automatically loaded by providing a four-letter PDB ID or via free text search in protein and ligand names. Another option is to upload custom structures in PDB format (e.g. result files from docking or molecular dynamics software). Output Figure 2 shows the result page for a typical analysis. For each binding site with a ligand, PLIP offers 2D and 3D in- teraction diagrams, a table with interaction details as well as downloadable result (XML and flat text) and visualization files (PNG and PyMOL session file). Details on interaction patterns can be accessed for each binding site by clicking on the identifier in the overview list. The results for each ligand are divided into a visualization section and a tabular listing of interaction data below (Figure 2). A JSMol-based 3D in- teraction diagram can be explored in the browser by clicking on the preview image. High-resolution images and PyMOL session files for preparation of custom publication-ready figures are available for download below the preview image. For manual inspection and successive processing, parsable Figure 2. PLIP result page. An interaction diagram and a table with inter- action data is provided for each binding site. JSMol applets allow to view XML or flat text files with interaction data are available at the 3D interaction diagrams in the browser. the bottom of the page. COMMAND-LINE TOOL DESCRIPTION structures, either directly from the PDB or in user-provided structures. Results for each binding site are provided as 3D The python source code of PLIP is available as open-source interaction diagrams for manual inspection (online in JS- software and allows to run computations locally. Additional mol and offline with PyMOL) as well as XML and flat text to the features of the web server, the PLIP command-line files for further processing. tool offers advanced settings for output files and thresholds. All interactions are listed on atom-level de- It enables high-throughput computation of protein struc- tail, enabling analyses of specific binding tures and can be readily integrated into analysis pipelines characteristics. PLIP is freely accessible at using the machine-readable result files. The usage and op- projects.biotec.tu-dresden.de/plip-web with- tions are explained in the Supplementary Data. out the need for registration or login. A short tutorial for new users as well as an extensive documentation is avail- PLIP ALGORITHM able on the website. The python source code is available for download on the PLIP website. Users interested in PLIP uses four steps to detect and report relevant in- batch processing are encouraged to use the tool locally in teractions: structure preparation, functional characteriza- command-line mode. A benchmark dataset of 30 literature- tion, rule-based matching and filtering of interactions (Fig- documented protein–ligand complexes is provided together ure 3A–D). The analysis is exemplified by Bacillus subtilis with the source code. DegV protein binding palmitic acid (PDB ID 3FYS). In the preparation step, the input structure is hydro- genated and ligands extracted along with their binding sites. WEB SERVER DESCRIPTION To this end, PLIP makes use of OpenBabel (21) for internal representation of molecules and most chemoinformatic cal- PLIP focuses on one-click processing of protein structures for the detection of interaction patterns. There are other culations. In order to retain only specifically binding small tools, web pages and databases (11–20)aswellassoft- molecules, PLIP uses a blacklist to exclude preparation ar- ware from Chemical Computing Group (MOE), Accelrys tifacts, modified residues, ions and solvent compounds as and CLC bio available. Many of those tools, however, are ligands. The full blacklist is available for download on the commercial or can be used for visualization purposes only. PLIP website. In the example, only palmitic acid is kept as Other offer only a limited selection of interaction types, re- a relevant ligand (Figure 3A). quire extensive preparation of input files or do not allow In order to find interacting groups (Figure 3B), the bind- processing of custom structures. With PLIP, comprehensive ing partners need to be functionally characterized first. interaction data for structures from PDB or external soft- This includes detection of hydrophobic atoms as well as ware is available without manual structure preparation and acceptors/donors for hydrogen and halogen bonds. Fur- is made available as both diagrams and parsable result files. thermore, PLIP searches for aromatic rings and charge cen- ters in protein and ligand. The latter functionalities are a precondition for formation of -stacking, -cation inter- Input actions or salt bridges. In the case of DegV with palmitic The user needs to provide a protein–ligand complex in PDB acid, charges can be assigned to two amino acids as well as format. Any structure from the RCSB PDB server (1)can the ligand carboxyl group (Figure 3B). Nucleic Acids Research, 2015, Vol. 43, Web Server issue W445 Figure 3. Example illustrating the four steps of PLIP in interaction detection for palmitic acid in Bacillus subtilis DegV protein (PDB ID 3FYS). (A) Structure preparation and detection of relevant ligands. (B) Functional characterization of molecules, here shown for the assignment of charges to amino acid side chains and the ligand carboxyl group. (C) Matching of interacting atoms using a rule-based system of geometric constraints. In the case of salt bridges, the distance between attracting charges is measured. (D) Filtering steps to minimize the number of depicted interactions, particularly important in the case of hydrophobic contacts (shown as solid gray lines). Following, putative interacting groups are matched by applying mostly geometric criteria (Figure 3C). Depending on the interaction type, this can include distance or angle constraints between arrangement of atoms. In the example case, the distances between atoms in positive and negative charges in the protein and palmitic acid are measured to de- cide whether to report a salt bridge (Figure 3C). The applied thresholds are taken from literature and are thus knowledge based. Most of them originate from analysis of large sets of high-quality protein structures in other studies. To ac- count for low-quality structures and structural errors, some thresholds have been modified to be more permissive. Last, filtering steps are used to eliminate redundant or overlapping interactions. As shown in Figure 3D this is especially important for hydrophobic contacts, which can be formed between any close apolar parts of ligand and protein. PLIP automatically searches for the most relevant contacts (shortest interatomic distance within the neigh- bourhood) to be reported. Some interaction types (e.g. salt bridges and hydrogen bonds) are very similar in their char- acteristics. In the case of detection of both interaction types for the same pairing of atoms, only one of them (e.g. a salt bridge) is reported. Detailed descriptions of the algorithm and thresholds for each interaction type are available as Supplementary Data. VALIDATION With the initial release of PLIP, we have included a test suite with 30 literature-validated examples (see Supplementary Data). They comprise diverse cases of protein–ligand com- plexes from PDB, covering all interaction types detectable by PLIP and resolutions from 1.2 to 3.3 A. For each case, a test was implemented to check whether all interactions re- Figure 4. Evaluating docking results with PLIP. Natural (A) and alterna- tive pose from redocking (B) of Cathepsin K with a small molecule in- ported in the corresponding paper are being detected. The hibitor (PDB ID 1VSN). Shared interacting residues are labeled. The sec- standard thresholds of PLIP have been carefully adapted to ond pose lacks characteristic halogen bonds. account for a broad range of interaction geometries while keeping the values as restrictive as possible. The test suite is available together with the source code on the PLIP website. W446 Nucleic Acids Research, 2015, Vol. 43, Web Server issue Figure 5. Human aldose reductase with different inhibitors. (A) Zenarestat (1IEI), (B) a sulfonyl-pyridazone inhibitor (1Z89) and (C) a benzothiazepine inhibitor (3P2V). While the first and the last share a salt bridge to His110 and the H-Bonds to Tyr48 and Trp111, there is a common -stacking to Trp111 in the first two. Unique interactions are, among others, two halogen bonds in zenarestat to the backbones of Val47 and Cys298, additional stacking with Trp20 for the benzothiazepine inhibitor and a water bridge to Trp20 in the last inhibitor. Large parts of all ligands bind via hydrophobic contacts. Example 2: inhibitor design Users are recommended to use these cases for testing when using custom thresholds and encouraged to contribute ad- In initial stages of inhibitor design or prior to library screen- ditional examples. ing, comparative analyses of known binding patterns with the target protein help identifying key residues. Here, PLIP is used to analyse interactions in three complexes of differ- EXAMPLES ent inhibitors (PDB IDs 1IEI, 1Z89, 3P2V) with human al- dose reductase (Figure 5). PLIP can be used for both––structures from the PDB Aldose reductase binds ligands via induced fit, leading to archive and structure files from other tools. It is therefore drastic conformational changes around the binding pocket possible to integrate PLIP into pipelines for analyses related (23). The three considered inhibitors show common in- to protein–ligand binding, e.g. post processing of docking teraction patterns but also individual subpatterns. While results or inhibitor design. both––zenarastat (Figure 5A) and the benzothiazepine in- hibitor (Figure 5C)––form a salt bridge to His110 and a hy- drogen bond to Trp111 via their carboxyl groups, the in- Example 1: docking post processing teraction pattern of the sulfonyl-pyridazone inhibitor (Fig- ure 5B) lacks this interaction. Without the carboxyl group The elimination of false positive results from docking re- only one hydrogen bond to Tyr48 is formed. This inter- sults can be performed using post processing pipelines (22). action can also be observed in complex with the benzoth- One approach is to use existing knowledge on key in- iazepine inhibitor. teractions with the protein of interest in order to filter Although all inhibitors have aromatic rings, only two from high-scoring poses. Cathepsin K in complex with a form -stacking interactions with Trp111. One of the most small molecule inhibitor (PDB ID 1VSN) was used for unique interaction patterns can be seen in the complex with a redocking experiment using the SwissDock server at zenarestat, where halogen bonds to the protein backbone swissdock.ch. While the top prediction corresponds to the are formed from both ends of the inhibitor. pose found in the crystal structure, the first alternative pose shows a clearly different ligand conformation, but compa- rable SwissDock fitness scores. PLIP was used to analyze the interaction patterns in the complex from the crystal structure (Figure 4A) and the complex with the alternative CONCLUSION pose (Figure 4B) from docking. In the alternative pose, the ligand part containing the aro- PLIP is the first web service to provide comprehensive anal- matic rings is flipped to the opposite direction. A rich net- ysis and visualization of non-covalent protein–ligand in- work of hydrogen bonds and water bridges can only be ob- teractions with one-click loading of structures. With the served for the correct pose (Figure 4A). Most strikingly, availability of PyMOL session files and results in parsable however, the characteristic halogen bonds are completely formats, both––manual inspection and computational pro- missing in the alternative pose, leaving the trifluoride group cessing of interaction data––are possible. Furthermore, the exposed. With the detailed interaction patterns at hand, it availability of PLIP source code enables local batch process- is thus possible to identify wrong poses based on previous ing, customization of the algorithm for special applications knowledge. as well as active development of the tool in the community. Nucleic Acids Research, 2015, Vol. 43, Web Server issue W447 SUPPLEMENTARY DATA 10. Haupt,V.J. and Schroeder,M. (2011) Old friends in new guise: repositioning of known drugs with structural bioinformatics. Brief Supplementary Data are available at NAR online. Bioinform., 12, 312–326. 11. Durrant,J. and McCammon,J. (2011) BINANA: a novel algorithm for ligand-binding characterization. J. Mol. Graph. Mod., 29, FUNDING 888–893. 12. Kasahara,K. and Kinoshita,K. (2014) GIANT: pattern analysis of GeneCloud and Hybris. Funding for open access molecular interactions in 3D structures of protein-small ligand charge: Public funding by Federal Government of Ger- complexes. BMC Bioinformatics, 15, 12. 13. Sobolev,V., Sorokine,a., Prilusky,J., Abola,E.E. and Edelman,M. many. (1999) Automated analysis of interatomic contacts in proteins. Conflict of interest statement. None declared. Bioinformatics, 15, 327–332. 14. Schreyer,A. and Blundell,T. (2009) CREDO: a protein-ligand interaction database for drug discovery. Chem. Biol. Drug Des., 73, REFERENCES 157–167. 15. Gallina,A.M., Bisignano,P., Bergamino,M. and Bordo,D. (2013) 1. Berman,H.M. (2008) The Protein Data Bank: a historical PLI: a web-based tool for the comparison of protein-ligand perspective. Acta Crystallogr. A, 64, 88–95. interactions observed on PDB structures. Bioinformatics, 29, 2. Konc,J. and Janezi ˇ c,D ˇ . (2014) ProBiS-ligands: a web server for 395–397. prediction of ligands by examination of protein binding sites. Nucleic 16. Hendlich,M., Bergner,A., Gunther,J ¨ . and Klebe,G. (2003) Relibase: Acids Res., 42, W215–W220. design and development of a database for comprehensive analysis of 3. Desaphy,J., Raimbaud,E., Ducrot,P. and Rognan,D. (2013) protein–ligand interactions. J. Mol. Biol., 326, 607–620. Encoding protein-ligand interaction patterns in fingerprints and 17. de Beer,T.A.P., Berka,K., Thornton,J.M. and Laskowski,R.a. (2014) graphs. J. Chem. Inf. Model., 53, 623–637. PDBsum additions. Nucleic Acids Res., 42, D292–D296. 4. Liu,Q., Kwoh,C.K. and Li,J. (2013) Binding affinity prediction for 18. Laskowski,R.A. and Swindells,M.B. (2011) LigPlot+: multiple protein-ligand complexes based on beta contacts and B factor. J. ligand-protein interaction diagrams for drug discovery. J. Chem. Inf. Chem. Inf. Model., 53, 3076–3085. Model., 51, 2778–2786. 5. Salentin,S., Haupt,V.J., Daminelli,S. and Schroeder,M. (2014) 19. Weisel,M., Bitter,H.-M., Diederich,F., So,W.V. and Kondru,R. Polypharmacology rescored: protein-ligand interaction profiles for (2012) PROLIX: rapid mining of protein-ligand interactions in large remote binding site similarity assessment. Prog. Biophys. Mol. Biol., crystal structure databases. J. Chem. Inf. Mod., 52, 1450–1461. 116, 174–186. 20. Clark,A.M. and Labute,P. (2007) 2D depiction of protein-ligand 6. Grosdidier,A., Zoete,V. and Michielin,O. (2011) SwissDock, a complexes. J. Chem. Inf. Mod., 47, 1933–1944. protein-small molecule docking web service based on EADock DSS. 21. O’Boyle,N.M., Morley,C. and Hutchison,G.R. (2008) Pybel: a Nucleic Acids Res., 39, W270–W277. Python wrapper for the OpenBabel cheminformatics toolkit. Chem. 7. Heo,L., Shin,W.-H., Lee,M.S. and Seok,C. (2014) GalaxySite: Cent. J., 2,5. ligand-binding-site prediction by using molecular docking. Nucleic 22. Yuriev,E. and Ramsland,P.A. (2013) Latest developments in Acids Res., 42, W210–W214. molecular docking: 2010-2011 in review. J. Mol. Recognit., 26, 8. Konc,J., Cesnik,T., Konc,J.T., Penca,M. and Janezi ˇ c,D ˇ . (2012) 215–239. ProBiS-database: precalculated binding site similarities and local 23. Koch,C., Heine,A. and Klebe,G. (2011) Ligand-induced fit affects pairwise alignments of PDB structures. J. Chem. Inf. Model., 52, binding modes and provokes changes in crystal packing of aldose 604–612. reductase. Biochim. Biophys. Acta, 1810, 879–887. 9. Haupt,V.J., Daminelli,S. and Schroeder,M. (2013) Drug promiscuity in PDB: protein binding site similarity is key. PLoS One, 8, e65894. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nucleic Acids Research Oxford University Press

PLIP: fully automated proteinligand interaction profiler

Loading next page...
 
/lp/oxford-university-press/plip-fully-automated-proteinligand-interaction-profiler-7l8gUlgzTP

References (23)

Publisher
Oxford University Press
Copyright
The Author(s) 2015. Published by Oxford University Press on behalf of Nucleic Acids Research.
ISSN
0305-1048
eISSN
1362-4962
DOI
10.1093/nar/gkv315
pmid
25873628
Publisher site
See Article on Publisher Site

Abstract

Published online 14 April 2015 Nucleic Acids Research, 2015, Vol. 43, Web Server issue W443–W447 doi: 10.1093/nar/gkv315 PLIP: fully automated protein–ligand interaction profiler 1 1 1 1,2 Sebastian Salentin , Sven Schreiber , V. Joachim Haupt , Melissa F. Adasme and 1,* Michael Schroeder 1 2 Biotechnology Center (BIOTEC), TU Dresden, Tatzberg 47-49, 01307 Dresden, Germany and Escuela de Ingenier´ ıa en Bioinformatica, ´ Universidad de Talca, Avda. Lircay s/n Talca, 3460000, Chile Received January 29, 2015; Revised March 17, 2015; Accepted March 28, 2015 ABSTRACT The characterization of interactions in protein–ligand complexes is essential for research in structural bioinformatics, drug discovery and biology. How- ever, comprehensive tools are not freely avail- able to the research community. Here, we present the protein–ligand interaction profiler (PLIP), a novel web service for fully automated detection and visualization of relevant non-covalent protein– ligand contacts in 3D structures, freely available at projects.biotec.tu-dresden.de/plip-web. The input is either a Protein Data Bank structure, a protein or ligand name, or a custom protein–ligand com- plex (e.g. from docking). In contrast to other tools, Figure 1. Example of interaction diagram generated with PLIP: Vari- the rule-based PLIP algorithm does not require any cella zoster virus thymidine kinase (1OSN) binding the antiherpes drug structure preparation. It returns a list of detected in- brivudine-monophosphate. The binding is dominated by a double - stacking and polar interactions at the terminal regions of the ligand. teractions on single atom level, covering seven in- teraction types (hydrogen bonds, hydrophobic con- tacts, pi-stacking, pi-cation interactions, salt bridges, lar recognition and protein function or to develop and op- water bridges and halogen bonds). PLIP stands out timize lead compounds. On the other hand, comparative by offering publication-ready images, PyMOL ses- high-throughput analyses of interaction patterns can con- sion files to generate custom images and parsable siderably improve protein–ligand docking or virtual screen- result files to facilitate successive data processing. ing (4) and thus enhance in silico approaches in drug discov- The full python source code is available for download ery. However, the scientific community lacks freely available on the website. PLIP’s command-line mode allows for tools to detect frequent non-covalent protein–ligand inter- high-throughput interaction profiling. actions such as hydrophobic interactions, hydrogen bonds, salt bridges and -stacking (5). To this end, we herein present the free and open-source protein–ligand interac- INTRODUCTION tion profiler (PLIP), a fully automated and easy to use web The Protein Data Bank (PDB) (1) hosts nearly 100 000 de- server and command-line tool for protein–ligand interac- posited protein structures, with over 75% of them solved tion detection (Figure 1). in complex with a small molecule ligand. Binding of a PLIP is complementary to other state-of-the-art web ligand to its host protein requires a specific arrangement tools such as SwissDock (6), GalaxySite (7)orProBiS(8) of attractive, typically non-covalent contacts between both and can thus be applied in evaluation of docking results molecules. With such rich data at hand, we can gain deep (Figure 4), drug design (Figure 5), binding site similarity insights into how ligands interact with their protein targets assessment (3,9) and drug repositioning (10). The PLIP (2,3). Detailed characterization of these interaction pat- web service allows for comprehensive detection and visu- terns in individual cases is crucial to understand molecu- alization of protein–ligand interaction patterns from 3D To whom correspondence should be addressed. Tel: +49 351 46340060; Fax: +49 351 46340061; Email: ms@biotec.tu-dresden.de C The Author(s) 2015. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. W444 Nucleic Acids Research, 2015, Vol. 43, Web Server issue be automatically loaded by providing a four-letter PDB ID or via free text search in protein and ligand names. Another option is to upload custom structures in PDB format (e.g. result files from docking or molecular dynamics software). Output Figure 2 shows the result page for a typical analysis. For each binding site with a ligand, PLIP offers 2D and 3D in- teraction diagrams, a table with interaction details as well as downloadable result (XML and flat text) and visualization files (PNG and PyMOL session file). Details on interaction patterns can be accessed for each binding site by clicking on the identifier in the overview list. The results for each ligand are divided into a visualization section and a tabular listing of interaction data below (Figure 2). A JSMol-based 3D in- teraction diagram can be explored in the browser by clicking on the preview image. High-resolution images and PyMOL session files for preparation of custom publication-ready figures are available for download below the preview image. For manual inspection and successive processing, parsable Figure 2. PLIP result page. An interaction diagram and a table with inter- action data is provided for each binding site. JSMol applets allow to view XML or flat text files with interaction data are available at the 3D interaction diagrams in the browser. the bottom of the page. COMMAND-LINE TOOL DESCRIPTION structures, either directly from the PDB or in user-provided structures. Results for each binding site are provided as 3D The python source code of PLIP is available as open-source interaction diagrams for manual inspection (online in JS- software and allows to run computations locally. Additional mol and offline with PyMOL) as well as XML and flat text to the features of the web server, the PLIP command-line files for further processing. tool offers advanced settings for output files and thresholds. All interactions are listed on atom-level de- It enables high-throughput computation of protein struc- tail, enabling analyses of specific binding tures and can be readily integrated into analysis pipelines characteristics. PLIP is freely accessible at using the machine-readable result files. The usage and op- projects.biotec.tu-dresden.de/plip-web with- tions are explained in the Supplementary Data. out the need for registration or login. A short tutorial for new users as well as an extensive documentation is avail- PLIP ALGORITHM able on the website. The python source code is available for download on the PLIP website. Users interested in PLIP uses four steps to detect and report relevant in- batch processing are encouraged to use the tool locally in teractions: structure preparation, functional characteriza- command-line mode. A benchmark dataset of 30 literature- tion, rule-based matching and filtering of interactions (Fig- documented protein–ligand complexes is provided together ure 3A–D). The analysis is exemplified by Bacillus subtilis with the source code. DegV protein binding palmitic acid (PDB ID 3FYS). In the preparation step, the input structure is hydro- genated and ligands extracted along with their binding sites. WEB SERVER DESCRIPTION To this end, PLIP makes use of OpenBabel (21) for internal representation of molecules and most chemoinformatic cal- PLIP focuses on one-click processing of protein structures for the detection of interaction patterns. There are other culations. In order to retain only specifically binding small tools, web pages and databases (11–20)aswellassoft- molecules, PLIP uses a blacklist to exclude preparation ar- ware from Chemical Computing Group (MOE), Accelrys tifacts, modified residues, ions and solvent compounds as and CLC bio available. Many of those tools, however, are ligands. The full blacklist is available for download on the commercial or can be used for visualization purposes only. PLIP website. In the example, only palmitic acid is kept as Other offer only a limited selection of interaction types, re- a relevant ligand (Figure 3A). quire extensive preparation of input files or do not allow In order to find interacting groups (Figure 3B), the bind- processing of custom structures. With PLIP, comprehensive ing partners need to be functionally characterized first. interaction data for structures from PDB or external soft- This includes detection of hydrophobic atoms as well as ware is available without manual structure preparation and acceptors/donors for hydrogen and halogen bonds. Fur- is made available as both diagrams and parsable result files. thermore, PLIP searches for aromatic rings and charge cen- ters in protein and ligand. The latter functionalities are a precondition for formation of -stacking, -cation inter- Input actions or salt bridges. In the case of DegV with palmitic The user needs to provide a protein–ligand complex in PDB acid, charges can be assigned to two amino acids as well as format. Any structure from the RCSB PDB server (1)can the ligand carboxyl group (Figure 3B). Nucleic Acids Research, 2015, Vol. 43, Web Server issue W445 Figure 3. Example illustrating the four steps of PLIP in interaction detection for palmitic acid in Bacillus subtilis DegV protein (PDB ID 3FYS). (A) Structure preparation and detection of relevant ligands. (B) Functional characterization of molecules, here shown for the assignment of charges to amino acid side chains and the ligand carboxyl group. (C) Matching of interacting atoms using a rule-based system of geometric constraints. In the case of salt bridges, the distance between attracting charges is measured. (D) Filtering steps to minimize the number of depicted interactions, particularly important in the case of hydrophobic contacts (shown as solid gray lines). Following, putative interacting groups are matched by applying mostly geometric criteria (Figure 3C). Depending on the interaction type, this can include distance or angle constraints between arrangement of atoms. In the example case, the distances between atoms in positive and negative charges in the protein and palmitic acid are measured to de- cide whether to report a salt bridge (Figure 3C). The applied thresholds are taken from literature and are thus knowledge based. Most of them originate from analysis of large sets of high-quality protein structures in other studies. To ac- count for low-quality structures and structural errors, some thresholds have been modified to be more permissive. Last, filtering steps are used to eliminate redundant or overlapping interactions. As shown in Figure 3D this is especially important for hydrophobic contacts, which can be formed between any close apolar parts of ligand and protein. PLIP automatically searches for the most relevant contacts (shortest interatomic distance within the neigh- bourhood) to be reported. Some interaction types (e.g. salt bridges and hydrogen bonds) are very similar in their char- acteristics. In the case of detection of both interaction types for the same pairing of atoms, only one of them (e.g. a salt bridge) is reported. Detailed descriptions of the algorithm and thresholds for each interaction type are available as Supplementary Data. VALIDATION With the initial release of PLIP, we have included a test suite with 30 literature-validated examples (see Supplementary Data). They comprise diverse cases of protein–ligand com- plexes from PDB, covering all interaction types detectable by PLIP and resolutions from 1.2 to 3.3 A. For each case, a test was implemented to check whether all interactions re- Figure 4. Evaluating docking results with PLIP. Natural (A) and alterna- tive pose from redocking (B) of Cathepsin K with a small molecule in- ported in the corresponding paper are being detected. The hibitor (PDB ID 1VSN). Shared interacting residues are labeled. The sec- standard thresholds of PLIP have been carefully adapted to ond pose lacks characteristic halogen bonds. account for a broad range of interaction geometries while keeping the values as restrictive as possible. The test suite is available together with the source code on the PLIP website. W446 Nucleic Acids Research, 2015, Vol. 43, Web Server issue Figure 5. Human aldose reductase with different inhibitors. (A) Zenarestat (1IEI), (B) a sulfonyl-pyridazone inhibitor (1Z89) and (C) a benzothiazepine inhibitor (3P2V). While the first and the last share a salt bridge to His110 and the H-Bonds to Tyr48 and Trp111, there is a common -stacking to Trp111 in the first two. Unique interactions are, among others, two halogen bonds in zenarestat to the backbones of Val47 and Cys298, additional stacking with Trp20 for the benzothiazepine inhibitor and a water bridge to Trp20 in the last inhibitor. Large parts of all ligands bind via hydrophobic contacts. Example 2: inhibitor design Users are recommended to use these cases for testing when using custom thresholds and encouraged to contribute ad- In initial stages of inhibitor design or prior to library screen- ditional examples. ing, comparative analyses of known binding patterns with the target protein help identifying key residues. Here, PLIP is used to analyse interactions in three complexes of differ- EXAMPLES ent inhibitors (PDB IDs 1IEI, 1Z89, 3P2V) with human al- dose reductase (Figure 5). PLIP can be used for both––structures from the PDB Aldose reductase binds ligands via induced fit, leading to archive and structure files from other tools. It is therefore drastic conformational changes around the binding pocket possible to integrate PLIP into pipelines for analyses related (23). The three considered inhibitors show common in- to protein–ligand binding, e.g. post processing of docking teraction patterns but also individual subpatterns. While results or inhibitor design. both––zenarastat (Figure 5A) and the benzothiazepine in- hibitor (Figure 5C)––form a salt bridge to His110 and a hy- drogen bond to Trp111 via their carboxyl groups, the in- Example 1: docking post processing teraction pattern of the sulfonyl-pyridazone inhibitor (Fig- ure 5B) lacks this interaction. Without the carboxyl group The elimination of false positive results from docking re- only one hydrogen bond to Tyr48 is formed. This inter- sults can be performed using post processing pipelines (22). action can also be observed in complex with the benzoth- One approach is to use existing knowledge on key in- iazepine inhibitor. teractions with the protein of interest in order to filter Although all inhibitors have aromatic rings, only two from high-scoring poses. Cathepsin K in complex with a form -stacking interactions with Trp111. One of the most small molecule inhibitor (PDB ID 1VSN) was used for unique interaction patterns can be seen in the complex with a redocking experiment using the SwissDock server at zenarestat, where halogen bonds to the protein backbone swissdock.ch. While the top prediction corresponds to the are formed from both ends of the inhibitor. pose found in the crystal structure, the first alternative pose shows a clearly different ligand conformation, but compa- rable SwissDock fitness scores. PLIP was used to analyze the interaction patterns in the complex from the crystal structure (Figure 4A) and the complex with the alternative CONCLUSION pose (Figure 4B) from docking. In the alternative pose, the ligand part containing the aro- PLIP is the first web service to provide comprehensive anal- matic rings is flipped to the opposite direction. A rich net- ysis and visualization of non-covalent protein–ligand in- work of hydrogen bonds and water bridges can only be ob- teractions with one-click loading of structures. With the served for the correct pose (Figure 4A). Most strikingly, availability of PyMOL session files and results in parsable however, the characteristic halogen bonds are completely formats, both––manual inspection and computational pro- missing in the alternative pose, leaving the trifluoride group cessing of interaction data––are possible. Furthermore, the exposed. With the detailed interaction patterns at hand, it availability of PLIP source code enables local batch process- is thus possible to identify wrong poses based on previous ing, customization of the algorithm for special applications knowledge. as well as active development of the tool in the community. Nucleic Acids Research, 2015, Vol. 43, Web Server issue W447 SUPPLEMENTARY DATA 10. Haupt,V.J. and Schroeder,M. (2011) Old friends in new guise: repositioning of known drugs with structural bioinformatics. Brief Supplementary Data are available at NAR online. Bioinform., 12, 312–326. 11. Durrant,J. and McCammon,J. (2011) BINANA: a novel algorithm for ligand-binding characterization. J. Mol. Graph. Mod., 29, FUNDING 888–893. 12. Kasahara,K. and Kinoshita,K. (2014) GIANT: pattern analysis of GeneCloud and Hybris. Funding for open access molecular interactions in 3D structures of protein-small ligand charge: Public funding by Federal Government of Ger- complexes. BMC Bioinformatics, 15, 12. 13. Sobolev,V., Sorokine,a., Prilusky,J., Abola,E.E. and Edelman,M. many. (1999) Automated analysis of interatomic contacts in proteins. Conflict of interest statement. None declared. Bioinformatics, 15, 327–332. 14. Schreyer,A. and Blundell,T. (2009) CREDO: a protein-ligand interaction database for drug discovery. Chem. Biol. Drug Des., 73, REFERENCES 157–167. 15. Gallina,A.M., Bisignano,P., Bergamino,M. and Bordo,D. (2013) 1. Berman,H.M. (2008) The Protein Data Bank: a historical PLI: a web-based tool for the comparison of protein-ligand perspective. Acta Crystallogr. A, 64, 88–95. interactions observed on PDB structures. Bioinformatics, 29, 2. Konc,J. and Janezi ˇ c,D ˇ . (2014) ProBiS-ligands: a web server for 395–397. prediction of ligands by examination of protein binding sites. Nucleic 16. Hendlich,M., Bergner,A., Gunther,J ¨ . and Klebe,G. (2003) Relibase: Acids Res., 42, W215–W220. design and development of a database for comprehensive analysis of 3. Desaphy,J., Raimbaud,E., Ducrot,P. and Rognan,D. (2013) protein–ligand interactions. J. Mol. Biol., 326, 607–620. Encoding protein-ligand interaction patterns in fingerprints and 17. de Beer,T.A.P., Berka,K., Thornton,J.M. and Laskowski,R.a. (2014) graphs. J. Chem. Inf. Model., 53, 623–637. PDBsum additions. Nucleic Acids Res., 42, D292–D296. 4. Liu,Q., Kwoh,C.K. and Li,J. (2013) Binding affinity prediction for 18. Laskowski,R.A. and Swindells,M.B. (2011) LigPlot+: multiple protein-ligand complexes based on beta contacts and B factor. J. ligand-protein interaction diagrams for drug discovery. J. Chem. Inf. Chem. Inf. Model., 53, 3076–3085. Model., 51, 2778–2786. 5. Salentin,S., Haupt,V.J., Daminelli,S. and Schroeder,M. (2014) 19. Weisel,M., Bitter,H.-M., Diederich,F., So,W.V. and Kondru,R. Polypharmacology rescored: protein-ligand interaction profiles for (2012) PROLIX: rapid mining of protein-ligand interactions in large remote binding site similarity assessment. Prog. Biophys. Mol. Biol., crystal structure databases. J. Chem. Inf. Mod., 52, 1450–1461. 116, 174–186. 20. Clark,A.M. and Labute,P. (2007) 2D depiction of protein-ligand 6. Grosdidier,A., Zoete,V. and Michielin,O. (2011) SwissDock, a complexes. J. Chem. Inf. Mod., 47, 1933–1944. protein-small molecule docking web service based on EADock DSS. 21. O’Boyle,N.M., Morley,C. and Hutchison,G.R. (2008) Pybel: a Nucleic Acids Res., 39, W270–W277. Python wrapper for the OpenBabel cheminformatics toolkit. Chem. 7. Heo,L., Shin,W.-H., Lee,M.S. and Seok,C. (2014) GalaxySite: Cent. J., 2,5. ligand-binding-site prediction by using molecular docking. Nucleic 22. Yuriev,E. and Ramsland,P.A. (2013) Latest developments in Acids Res., 42, W210–W214. molecular docking: 2010-2011 in review. J. Mol. Recognit., 26, 8. Konc,J., Cesnik,T., Konc,J.T., Penca,M. and Janezi ˇ c,D ˇ . (2012) 215–239. ProBiS-database: precalculated binding site similarities and local 23. Koch,C., Heine,A. and Klebe,G. (2011) Ligand-induced fit affects pairwise alignments of PDB structures. J. Chem. Inf. Model., 52, binding modes and provokes changes in crystal packing of aldose 604–612. reductase. Biochim. Biophys. Acta, 1810, 879–887. 9. Haupt,V.J., Daminelli,S. and Schroeder,M. (2013) Drug promiscuity in PDB: protein binding site similarity is key. PLoS One, 8, e65894.

Journal

Nucleic Acids ResearchOxford University Press

Published: Jul 1, 2015

There are no references for this article.