Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

ChemoPy: freely available python package for computational biology and chemoinformatics

ChemoPy: freely available python package for computational biology and chemoinformatics Vol. 29 no. 8 2013, pages 1092–1094 BIOINFORMATICS APPLICATIONS NOTE doi:10.1093/bioinformatics/btt105 Systems biology Advance Access publication March 14, 2013 ChemoPy: freely available python package for computational biology and chemoinformatics 1 2 3 1, Dong-Sheng Cao ,Qing-Song Xu ,Qian-Nan Hu and Yi-Zeng Liang Research Center of Modernization of Traditional Chinese Medicines, Central South University, Changsha 410083, P. R. China, School of Mathematics and Statistics, Central South University, Changsha 410083, P. R. China and Key Laboratory of Combinatorial Biosynthesis and Drug Discovery (Wuhan University), Ministry of Education, Wuhan 430071, P. R. China Associate Editor: Martin Bishop 2008; Cao et al., 2012a, b; Chou et al., 2006; Izrailev et al., 2004; ABSTRACT Keiser et al., 2007; Prado-Prado et al., 2011a, b; Vin˜ a et al., Motivation: Molecular representation for small molecules has been 2009), following the spirit of chemogenomics. routinely used in QSAR/SAR, virtual screening, database search, rank- Several programs for computing molecular descriptors have ing, drug ADME/T prediction and other drug discovery processes. To been developed, such as MARCH-INSIDE, TOPS-MODE, facilitate extensive studies of drug molecules, we developed a freely TOMO-COMD, Dragon, CODESSA, Molconn-Z (http://www. available, open-source python package called chemoinformatics in edusoft-lc.com/molconn/), Chemistry Development Kit (CDK), python (ChemoPy) for calculating the commonly used structural and Indigo (http://ggasoftware.com/opensource/indigo), JOELib, physicochemical features. It computes 16 drug feature groups com- RDKit (http://www.rdkit.org/) and Avogadro (Gonza´ lez-Dı´az posed of 19 descriptors that include 1135 descriptor values. In add- et al., 2008; Hanwell et al., 2012; Katritzky et al., 1994; ition, it provides seven types of molecular fingerprint systems for drug Marrero-Ponce et al., 2002; Pe´ rez-Gonza´ lez et al., 2003; molecules, including topological fingerprints, electro-topological state Steinbeck et al., 2003; Todeschini et al., 2010; Wegner, 2005). (E-state) fingerprints, MACCS keys, FP4 keys, atom pairs fingerprints, Unfortunately, some of these tools are not comprehensive, or topological torsion fingerprints and Morgan/circular fingerprints. By are limited to only a certain kind of features. Additionally, applying a semi-empirical quantum chemistry program MOPAC, some are not freely and easily accessible. ChemoPy can also compute a large number of 3D molecular descrip- We implemented a selection of sophisticated molecular features tors conveniently. and provided them as a package for the free and open-source Availability: The python package, ChemoPy, is freely available via software environment python. The ChemoPy package aims at http://code.google.com/p/pychem/downloads/list, and it runs on providing the user with comprehensive implementations of these Linux and MS-Windows. descriptors in a unified framework to allow easy and transparent Contact: [email protected] computation. To our knowledge, ChemoPy is the first open- Supplementary information: Supplementary data are available at source package computing a large number of molecular features Bioinformatics online. based on the MOPAC optimization. We recommend ChemoPy to Received on November 27, 2012; revised on February 22, 2013; analyse and represent the drugs or ligand molecules under inves- accepted on February 25, 2013 tigation. Further, we hope that the package will be helpful when exploring questions concerning drug activity, drug ADME/T and drug–target interactions in the context of computational biology. 1INTRODUCTION After accomplishing the previous goal, we expect that our/other Molecular features for small molecules have frequently been used groups may use the free code of ChemoPy and the new QSAR in the development of machine learning in QSAR/QSPR, virtual models to implement public web servers, such as MIND-BEST screening, database search, similarity search, ranking, drug (Gonza´ lez-Dı´az et al., 2011). The users can run predictions of ADME/T prediction and other drug discovery processes (Cao libraries of compounds using SMILES codes as input. et al., 2010, 2011, 2012a, b; Dea-Ayuela et al., 2008; Du et al., ´ ´ 2005, 2008a, b, 2009; Gola et al., 2006, Gonzalez-Dıaz et al., 2 CHEMOPY FEATURES 2005; Prado-Prado et al., 2008, 2009, 2010; van de Waterbeemd et al., 2003; Wang et al., 2011; Wei et al., 2009; The ChemoPy package contains several functions and modules Yan et al., 2012; Zhu et al., 2011). These descriptors capture manipulating drug molecules. To obtain molecular structures and magnify distinct aspects of molecular topology to investigate easily, ChemoPy provides a download module, by which the how molecular structure affects molecular property. Currently, user could easily get molecular structures from four databases these features were widely used to characterize ligand molecules (i.e. KEGG, PubChem, DrugBank and CAS) by providing IDs. in the protein–ligand network and predict new protein–ligand 444 from pychem.pychem import getmol associations to identify potential drug targets (Campillos et al., 444 smi1 ¼ getmol.GetMolFromNCBI(‘2244’) *To whom correspondence should be addressed. 444 print smi1 1092  The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected] ChemoPy CC(Oc1ccccc1C(O)¼O)¼O The other is to call the PyChem2d or PyChem3d class by im- porting the pychem module, which encapsulates commonly used 444 smi2 ¼ getmol.GetMolFromKegg(‘D00109’) descriptor calculation methods. PyChem2d and PyChem3d are 444 print smi2 responsible for the calculation of 2D and 3D molecular descrip- CC(Oc1ccccc1C(O)¼O)¼O tors, respectively. We could construct a PyChem2d or PyChem3d object with a molecule input and then call corres- ChemoPy can compute a large number of 2D and 3D descrip- ponding methods to calculate these features. tors. A list of structural and physicochemical features covered by ChemoPy is summarized in Table 1 (see also detailed descriptor 444 from pychem.pychem import PyChem2d, PyChem3d list in Supplementary Material S1). There are two means to com- 444 des1 ¼ PyChem2d() pute these molecular descriptors from small molecules. One is to 444 des1.ReadMolFromSmile(‘‘CC(Oc1ccccc1C(O)¼O)¼O’’) use the built-in modules. There exist 19 modules responding to the calculation of descriptors from 16 feature groups. The in- 444 AllConnectivity ¼ des1.GetConnectivity() struction for each module is provided in the form of HTML in 444 des2 ¼ PyChem3d() ChemoPy (see Supplementary Material S2). We could import 444 des2.ReadMol(‘‘CC(Oc1ccccc1C(O)¼O)¼O’’) related functions to compute these features. For example, the 444 All3D ¼ des2.GetAllDescriptor() topology module includes a number of functionalities used for calculating various topological descriptors. The user could con- In ChemoPy, molecular structures are optimized by the AM1 veniently use them as need. method in MOPAC. MOPAC input file is directly prepared by Pybel and OpenBabel. The detailed introductions for all descrip- 444 from pychem.pychem import Chem, topology tors are provided in the ChemoPy manual (see Supplementary 444 mol ¼ Chem.MolFromSmiles(‘‘CC(Oc1ccccc1C(O)¼O)¼O’’) Material S3). A user guide for the use of ChemoPy is included to 444 Weiner ¼ topology.CalculateWeiner(mol) guide how the user uses it to calculate the needed features (see 444 Alltopology ¼ topology.GetTopology(mol) Supplementary Material S4). ChemoPy is written by the pure python language. We chose to use python because it is open source, and there already exist packages to handle small molecules [e.g. Pybel (O’Boyle et al., 2008b), PyMol and Cinfony (O’Boyle et al., 2008a)]. It is con- Table 1. List of ChemoPy computed features for small molecules venient for ChemoPy to analyse drug molecules processed by Cinfony or RDKit. ChemoPy is available for two operating Feature group Features Number of systems: Linux and Windows. ChemoPy depends on Pybel, descriptors RDKit, OpenBabel (O’Boyle et al., 2011) and MOPAC (Stewart, 1990). Moreover, it needs the support of scientific Constitution Constitutional descriptors 30 library for python (SciPy). Topology Topological descriptors 35 Connectivity Connectivity indices 44 E-state E-state descriptors 245 Kappa Kappa shape descriptors 7 3DISCUSSION Basak Basak information indices 21 ChemoPy contains a selection of molecular descriptors to Burden Burden descriptors 64 analyse, classify and compare complex molecular network. Autocorrelation Moreau-Broto autocorrelation 32 Moran autocorrelation 32 They facilitate to exploit machine-learning techniques to drive Geary autocorrelation 32 hypothesis from complex molecular datasets. The usefulness of Charge Charge descriptors 25 these molecular descriptors covered by ChemoPy for represent- Property Molecular property 6 ing structural features of small molecules has been suffi- MOE-type MOE-type descriptors 60 ciently demonstrated by a number of published studies of the Geometric Geometric descriptors 12 development of machine-learning classification systems. The CPSA CPSA descriptors 30 RDF RDF descriptors 180 ChemoPy implementation of each of these algorithms was MoRSE MoRSE descriptors 210 extensively tested by using a number of test molecules. The com- WHIM WHIM descriptors 70 puted descriptor values were compared with the known Fingerprints Topological fingerprints 2048 b values for these molecules to ensure that our computation is MACCS keys 166 accurate. FP4 keys 307 Owing to the modular structure of ChemoPy, extensions or E-state fingerprints 79 new functionalities can be implemented easily without complex Atom pairs fingerprints — Topological torsions — and time-consuming alterations of the source code. In future Morgan fingerprints — work, we plan to apply the integrated features on various biolo- gical research questions, and extending the range of functions Some of these features are from RDKit. with new promising descriptors for the coming versions of These features are from RDKit. These features are from OpenBabel. ChemoPy. 1093 D.-S.Cao et al. Katritzky,A.R. et al. (1994) CODESSA Comprehensive Descriptors for Structural ACKNOWLEDGEMENTS and Statistical Analysis. Reference manual. The authors thank three anonymous referees for their construct- Keiser,M.J. et al. (2007) Relating protein pharmacology by ligand chemistry. Nat. Biotech., 25, 197–206. ive comments, which greatly helped improve on the original Marrero-Ponce,Y. et al. (2002) TOMOCOMD software, version 1.0, 2002.Central version of the manuscript. University of Las Villas. O’Boyle,N. et al. (2008a) Cinfony—combining open source cheminformatics tool- Funding: National Natural Science Foundation of China kits behind a common interface. Chem. Cent. J., 2,24. (21075138, 21275164 and 11271374). O’Boyle,N. et al. (2008b) Pybel: a Python wrapper for the OpenBabel cheminfor- matics toolkit. Chem. Cent. J., 2,1–5. Conflict of Interest: none declared. O’Boyle,N. et al. (2011) Open babel: an open chemical toolbox. J. Cheminform., 3, 1–14. Pe´ rez-Gonza´ lez,M. et al. (2003) TOPS-MODE based QSARs derived from hetero- REFERENCES geneous series of compounds. Applications to the design of new herbicides. J. Chem. Inf. Comput. Sci., 43, 1192–1199. Campillos,M. et al. (2008) Drug target identification using side-effect similarity. Prado-Prado,F.J. et al. (2008) Unified QSAR approach to antimicrobials. Part 3: Science, 321, 263–266. first multi-tasking QSAR model for input-coded prediction, structural Cao,D.S. et al. (2010) Prediction of aqueous solubility of druglike organic com- back-projection, and complex networks clustering of antiprotozoal compounds. pounds using partial least squares, back-propagation network and support Bioorg.Med.Chem., 16, 5871–5880. vector machine. J. Chemometr., 24, 584–595. Prado-Prado,F.J. et al. (2009) Unified QSAR approach to antimicrobials. 4. Cao,D.S. et al. (2011) In silico classification of human maximum recommended Multi-target QSAR modeling and comparative multi-distance study of the daily dose based on modified random forest and substructure fingerprint. giant components of antiviral drug-drug complex networks. Bioorg. Med. Anal. Chim. Acta., 692, 50–56. Chem., 17, 569–575. Cao,D.S. et al. (2012a) Computer-aided prediction of toxicity with substructure Prado-Prado,F.J. et al. (2010) Multi-target spectral moment QSAR versus ANN for pattern and random forest. J. Chemometr., 26,7–15. antiparasitic drugs against different parasite species. Bioorg. Med. Chem., 18, Cao,D.S. et al. (2012b) Large-scale prediction of drug-target interactions using pro- 2225–2231. tein sequences and drug topological structures. Anal. Chim. Acta., 752,1–10. Prado-Prado,F.J. et al. (2011a) Using entropy of drug and protein graphs to predict Chou,K.C. et al. (2006) Predicting networking couples for metabolic pathways of FDA drug-target network: theoretical-experimental study of MAO inhibitors Arabidopsis. EXCLI J., 5,55–65. and hemoglobin peptides from Fasciola hepatica. Eur. J. Med. Chem., 46, Dea-Ayuela,M.A. et al. (2008) HP-Lattice QSAR for dynein proteins: experimental 1074–1094. proteomics (2D-electrophoresis, mass spectrometry) and theoretic study of a Prado-Prado,F.J. et al. (2011b) 2D MI-DRAGON: a new predictor for Leishmania infantum sequence. Bioorg.Med.Chem., 16, 7770–7776. protein-ligands interactions and theoretic-experimental studies of US FDA Du,Q.S. et al. (2005) Heuristic molecular lipophilicity potential (HMLP): a drug-target network, oxoisoaporphine inhibitors for MAO-A and human para- 2D-QSAR study to LADH of molecular family pyrazole and derivatives. site proteins. Eur. J. Med. Chem., 46, 5838–5851. J. Comput. Chem., 26, 461–470. Steinbeck,C. et al. (2003) The chemistry development kit (CDK): an open-source Du,Q.S. et al. (2008a) Multiple field three dimensional quantitative java library for chemo- and bioinformatics. J. Chem. Inf. Comput. Sci., 43, structure-activity relationship (MF-3D-QSAR). J. Comput. Chem., 29, 211–219. 493–500. Du,Q.S. et al. (2008b) Recent advances in QSAR and their applications in predict- Stewart,J.P. (1990) MOPAC: a semiempirical molecular orbital program. ing the activities of chemical molecules, peptides and proteins for drug design. J. Comput. Aided Mol. Des., 4,1–103. Curr. Protein Pept. Sci., 9, 248–259. Todeschini,R. et al. (2010) Molecular Descriptors for Chemoinformatics.Wiley-VCH Du,Q.S. et al. (2009) Fragment-based quantitative structure-activity relationship GmbH & Co. KGaA, Weinheim. (FB-QSAR) for fragment-based drug design. J. Comput. Chem., 30, 295–304. van de Waterbeemd,H. et al. (2003) ADMET in silico modelling: towards prediction Gola,J. et al. (2006) ADMET property prediction: the state of the art and current paradise? Nat. Rev. Drug Discov., 2, 192–204. challenges. QSAR Comb. Sci., 25, 1172–1180. Vin˜ a,D. et al. (2009) Alingment-free prediction of a drug-target complex network Gonza´ lez-Dı´az,H. et al. (2005) Predicting multiple drugs side effects with a general based on parameters of drug connectivity and protein sequence of receptors. drug-target interaction thermodynamic Markov model. Bioorg.Med.Chem., 13, Mol. Pharm., 6, 825–835. 1119–1129. Wang,J.M. et al. (2011) Recent advances on aqueous solubility prediction. Comb. Gonza´ lez-Dı´az,H. et al. (2008) Predicting antimicrobial drugs and targets with the Chem. High Throughput Screen., 14, 328–338. MARCH-INSIDE approach. Curr. Top. Med. Chem., 8, 1676–1690. Wegner,J.K. (2005) JOELib: Graph/Data Mining and Clustering. ACS Meeting. Gonza´ lez-Dı´az,H. et al. (2011) MIND-BEST: web server for drugs and target dis- Wei,H. et al. (2009) Investigation into adamantane-based M2 inhibitors with covery; design, synthesis, and assay of MAO-B inhibitors and theoretical ex- FB-QSAR. Med. Chem., 5, 305–317. perimental study of G3PDH protein from Trichomonas gallinae. J. Proteome. Yan,J. et al. (2012) Comparison of quantitative structure-retention relationship Res., 10, 1698–1718. Hanwell,M.D. et al. (2012) Avogadro: an advanced semantic chemical editor, visu- models on four stationary phases with different polarity for a diverse set of alization, and analysis platform. J. Chemoinform., 4,17. flavor compounds. J. Chromatogr. A, 1223, 118–125. Izrailev,S. et al. (2004) Enzyme classification by ligand binding. Proteins, 57, Zhu,J.Y. et al. (2011) Recent developments of in silico predictions of oral bioavail- 711–724. ability. Comb. Chem. High Throughput Screen., 14, 362–374. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Bioinformatics Oxford University Press

ChemoPy: freely available python package for computational biology and chemoinformatics

Download PDF
3 pages

Loading next page...
 
/lp/oxford-university-press/chemopy-freely-available-python-package-for-computational-biology-and-wfZmaFli4G

References (44)

Publisher
Oxford University Press
Copyright
© The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]
ISSN
1367-4803
eISSN
1460-2059
DOI
10.1093/bioinformatics/btt105
pmid
23493324
Publisher site
See Article on Publisher Site

Abstract

Vol. 29 no. 8 2013, pages 1092–1094 BIOINFORMATICS APPLICATIONS NOTE doi:10.1093/bioinformatics/btt105 Systems biology Advance Access publication March 14, 2013 ChemoPy: freely available python package for computational biology and chemoinformatics 1 2 3 1, Dong-Sheng Cao ,Qing-Song Xu ,Qian-Nan Hu and Yi-Zeng Liang Research Center of Modernization of Traditional Chinese Medicines, Central South University, Changsha 410083, P. R. China, School of Mathematics and Statistics, Central South University, Changsha 410083, P. R. China and Key Laboratory of Combinatorial Biosynthesis and Drug Discovery (Wuhan University), Ministry of Education, Wuhan 430071, P. R. China Associate Editor: Martin Bishop 2008; Cao et al., 2012a, b; Chou et al., 2006; Izrailev et al., 2004; ABSTRACT Keiser et al., 2007; Prado-Prado et al., 2011a, b; Vin˜ a et al., Motivation: Molecular representation for small molecules has been 2009), following the spirit of chemogenomics. routinely used in QSAR/SAR, virtual screening, database search, rank- Several programs for computing molecular descriptors have ing, drug ADME/T prediction and other drug discovery processes. To been developed, such as MARCH-INSIDE, TOPS-MODE, facilitate extensive studies of drug molecules, we developed a freely TOMO-COMD, Dragon, CODESSA, Molconn-Z (http://www. available, open-source python package called chemoinformatics in edusoft-lc.com/molconn/), Chemistry Development Kit (CDK), python (ChemoPy) for calculating the commonly used structural and Indigo (http://ggasoftware.com/opensource/indigo), JOELib, physicochemical features. It computes 16 drug feature groups com- RDKit (http://www.rdkit.org/) and Avogadro (Gonza´ lez-Dı´az posed of 19 descriptors that include 1135 descriptor values. In add- et al., 2008; Hanwell et al., 2012; Katritzky et al., 1994; ition, it provides seven types of molecular fingerprint systems for drug Marrero-Ponce et al., 2002; Pe´ rez-Gonza´ lez et al., 2003; molecules, including topological fingerprints, electro-topological state Steinbeck et al., 2003; Todeschini et al., 2010; Wegner, 2005). (E-state) fingerprints, MACCS keys, FP4 keys, atom pairs fingerprints, Unfortunately, some of these tools are not comprehensive, or topological torsion fingerprints and Morgan/circular fingerprints. By are limited to only a certain kind of features. Additionally, applying a semi-empirical quantum chemistry program MOPAC, some are not freely and easily accessible. ChemoPy can also compute a large number of 3D molecular descrip- We implemented a selection of sophisticated molecular features tors conveniently. and provided them as a package for the free and open-source Availability: The python package, ChemoPy, is freely available via software environment python. The ChemoPy package aims at http://code.google.com/p/pychem/downloads/list, and it runs on providing the user with comprehensive implementations of these Linux and MS-Windows. descriptors in a unified framework to allow easy and transparent Contact: [email protected] computation. To our knowledge, ChemoPy is the first open- Supplementary information: Supplementary data are available at source package computing a large number of molecular features Bioinformatics online. based on the MOPAC optimization. We recommend ChemoPy to Received on November 27, 2012; revised on February 22, 2013; analyse and represent the drugs or ligand molecules under inves- accepted on February 25, 2013 tigation. Further, we hope that the package will be helpful when exploring questions concerning drug activity, drug ADME/T and drug–target interactions in the context of computational biology. 1INTRODUCTION After accomplishing the previous goal, we expect that our/other Molecular features for small molecules have frequently been used groups may use the free code of ChemoPy and the new QSAR in the development of machine learning in QSAR/QSPR, virtual models to implement public web servers, such as MIND-BEST screening, database search, similarity search, ranking, drug (Gonza´ lez-Dı´az et al., 2011). The users can run predictions of ADME/T prediction and other drug discovery processes (Cao libraries of compounds using SMILES codes as input. et al., 2010, 2011, 2012a, b; Dea-Ayuela et al., 2008; Du et al., ´ ´ 2005, 2008a, b, 2009; Gola et al., 2006, Gonzalez-Dıaz et al., 2 CHEMOPY FEATURES 2005; Prado-Prado et al., 2008, 2009, 2010; van de Waterbeemd et al., 2003; Wang et al., 2011; Wei et al., 2009; The ChemoPy package contains several functions and modules Yan et al., 2012; Zhu et al., 2011). These descriptors capture manipulating drug molecules. To obtain molecular structures and magnify distinct aspects of molecular topology to investigate easily, ChemoPy provides a download module, by which the how molecular structure affects molecular property. Currently, user could easily get molecular structures from four databases these features were widely used to characterize ligand molecules (i.e. KEGG, PubChem, DrugBank and CAS) by providing IDs. in the protein–ligand network and predict new protein–ligand 444 from pychem.pychem import getmol associations to identify potential drug targets (Campillos et al., 444 smi1 ¼ getmol.GetMolFromNCBI(‘2244’) *To whom correspondence should be addressed. 444 print smi1 1092  The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected] ChemoPy CC(Oc1ccccc1C(O)¼O)¼O The other is to call the PyChem2d or PyChem3d class by im- porting the pychem module, which encapsulates commonly used 444 smi2 ¼ getmol.GetMolFromKegg(‘D00109’) descriptor calculation methods. PyChem2d and PyChem3d are 444 print smi2 responsible for the calculation of 2D and 3D molecular descrip- CC(Oc1ccccc1C(O)¼O)¼O tors, respectively. We could construct a PyChem2d or PyChem3d object with a molecule input and then call corres- ChemoPy can compute a large number of 2D and 3D descrip- ponding methods to calculate these features. tors. A list of structural and physicochemical features covered by ChemoPy is summarized in Table 1 (see also detailed descriptor 444 from pychem.pychem import PyChem2d, PyChem3d list in Supplementary Material S1). There are two means to com- 444 des1 ¼ PyChem2d() pute these molecular descriptors from small molecules. One is to 444 des1.ReadMolFromSmile(‘‘CC(Oc1ccccc1C(O)¼O)¼O’’) use the built-in modules. There exist 19 modules responding to the calculation of descriptors from 16 feature groups. The in- 444 AllConnectivity ¼ des1.GetConnectivity() struction for each module is provided in the form of HTML in 444 des2 ¼ PyChem3d() ChemoPy (see Supplementary Material S2). We could import 444 des2.ReadMol(‘‘CC(Oc1ccccc1C(O)¼O)¼O’’) related functions to compute these features. For example, the 444 All3D ¼ des2.GetAllDescriptor() topology module includes a number of functionalities used for calculating various topological descriptors. The user could con- In ChemoPy, molecular structures are optimized by the AM1 veniently use them as need. method in MOPAC. MOPAC input file is directly prepared by Pybel and OpenBabel. The detailed introductions for all descrip- 444 from pychem.pychem import Chem, topology tors are provided in the ChemoPy manual (see Supplementary 444 mol ¼ Chem.MolFromSmiles(‘‘CC(Oc1ccccc1C(O)¼O)¼O’’) Material S3). A user guide for the use of ChemoPy is included to 444 Weiner ¼ topology.CalculateWeiner(mol) guide how the user uses it to calculate the needed features (see 444 Alltopology ¼ topology.GetTopology(mol) Supplementary Material S4). ChemoPy is written by the pure python language. We chose to use python because it is open source, and there already exist packages to handle small molecules [e.g. Pybel (O’Boyle et al., 2008b), PyMol and Cinfony (O’Boyle et al., 2008a)]. It is con- Table 1. List of ChemoPy computed features for small molecules venient for ChemoPy to analyse drug molecules processed by Cinfony or RDKit. ChemoPy is available for two operating Feature group Features Number of systems: Linux and Windows. ChemoPy depends on Pybel, descriptors RDKit, OpenBabel (O’Boyle et al., 2011) and MOPAC (Stewart, 1990). Moreover, it needs the support of scientific Constitution Constitutional descriptors 30 library for python (SciPy). Topology Topological descriptors 35 Connectivity Connectivity indices 44 E-state E-state descriptors 245 Kappa Kappa shape descriptors 7 3DISCUSSION Basak Basak information indices 21 ChemoPy contains a selection of molecular descriptors to Burden Burden descriptors 64 analyse, classify and compare complex molecular network. Autocorrelation Moreau-Broto autocorrelation 32 Moran autocorrelation 32 They facilitate to exploit machine-learning techniques to drive Geary autocorrelation 32 hypothesis from complex molecular datasets. The usefulness of Charge Charge descriptors 25 these molecular descriptors covered by ChemoPy for represent- Property Molecular property 6 ing structural features of small molecules has been suffi- MOE-type MOE-type descriptors 60 ciently demonstrated by a number of published studies of the Geometric Geometric descriptors 12 development of machine-learning classification systems. The CPSA CPSA descriptors 30 RDF RDF descriptors 180 ChemoPy implementation of each of these algorithms was MoRSE MoRSE descriptors 210 extensively tested by using a number of test molecules. The com- WHIM WHIM descriptors 70 puted descriptor values were compared with the known Fingerprints Topological fingerprints 2048 b values for these molecules to ensure that our computation is MACCS keys 166 accurate. FP4 keys 307 Owing to the modular structure of ChemoPy, extensions or E-state fingerprints 79 new functionalities can be implemented easily without complex Atom pairs fingerprints — Topological torsions — and time-consuming alterations of the source code. In future Morgan fingerprints — work, we plan to apply the integrated features on various biolo- gical research questions, and extending the range of functions Some of these features are from RDKit. with new promising descriptors for the coming versions of These features are from RDKit. These features are from OpenBabel. ChemoPy. 1093 D.-S.Cao et al. Katritzky,A.R. et al. (1994) CODESSA Comprehensive Descriptors for Structural ACKNOWLEDGEMENTS and Statistical Analysis. Reference manual. The authors thank three anonymous referees for their construct- Keiser,M.J. et al. (2007) Relating protein pharmacology by ligand chemistry. Nat. Biotech., 25, 197–206. ive comments, which greatly helped improve on the original Marrero-Ponce,Y. et al. (2002) TOMOCOMD software, version 1.0, 2002.Central version of the manuscript. University of Las Villas. O’Boyle,N. et al. (2008a) Cinfony—combining open source cheminformatics tool- Funding: National Natural Science Foundation of China kits behind a common interface. Chem. Cent. J., 2,24. (21075138, 21275164 and 11271374). O’Boyle,N. et al. (2008b) Pybel: a Python wrapper for the OpenBabel cheminfor- matics toolkit. Chem. Cent. J., 2,1–5. Conflict of Interest: none declared. O’Boyle,N. et al. (2011) Open babel: an open chemical toolbox. J. Cheminform., 3, 1–14. Pe´ rez-Gonza´ lez,M. et al. (2003) TOPS-MODE based QSARs derived from hetero- REFERENCES geneous series of compounds. Applications to the design of new herbicides. J. Chem. Inf. Comput. Sci., 43, 1192–1199. Campillos,M. et al. (2008) Drug target identification using side-effect similarity. Prado-Prado,F.J. et al. (2008) Unified QSAR approach to antimicrobials. Part 3: Science, 321, 263–266. first multi-tasking QSAR model for input-coded prediction, structural Cao,D.S. et al. (2010) Prediction of aqueous solubility of druglike organic com- back-projection, and complex networks clustering of antiprotozoal compounds. pounds using partial least squares, back-propagation network and support Bioorg.Med.Chem., 16, 5871–5880. vector machine. J. Chemometr., 24, 584–595. Prado-Prado,F.J. et al. (2009) Unified QSAR approach to antimicrobials. 4. Cao,D.S. et al. (2011) In silico classification of human maximum recommended Multi-target QSAR modeling and comparative multi-distance study of the daily dose based on modified random forest and substructure fingerprint. giant components of antiviral drug-drug complex networks. Bioorg. Med. Anal. Chim. Acta., 692, 50–56. Chem., 17, 569–575. Cao,D.S. et al. (2012a) Computer-aided prediction of toxicity with substructure Prado-Prado,F.J. et al. (2010) Multi-target spectral moment QSAR versus ANN for pattern and random forest. J. Chemometr., 26,7–15. antiparasitic drugs against different parasite species. Bioorg. Med. Chem., 18, Cao,D.S. et al. (2012b) Large-scale prediction of drug-target interactions using pro- 2225–2231. tein sequences and drug topological structures. Anal. Chim. Acta., 752,1–10. Prado-Prado,F.J. et al. (2011a) Using entropy of drug and protein graphs to predict Chou,K.C. et al. (2006) Predicting networking couples for metabolic pathways of FDA drug-target network: theoretical-experimental study of MAO inhibitors Arabidopsis. EXCLI J., 5,55–65. and hemoglobin peptides from Fasciola hepatica. Eur. J. Med. Chem., 46, Dea-Ayuela,M.A. et al. (2008) HP-Lattice QSAR for dynein proteins: experimental 1074–1094. proteomics (2D-electrophoresis, mass spectrometry) and theoretic study of a Prado-Prado,F.J. et al. (2011b) 2D MI-DRAGON: a new predictor for Leishmania infantum sequence. Bioorg.Med.Chem., 16, 7770–7776. protein-ligands interactions and theoretic-experimental studies of US FDA Du,Q.S. et al. (2005) Heuristic molecular lipophilicity potential (HMLP): a drug-target network, oxoisoaporphine inhibitors for MAO-A and human para- 2D-QSAR study to LADH of molecular family pyrazole and derivatives. site proteins. Eur. J. Med. Chem., 46, 5838–5851. J. Comput. Chem., 26, 461–470. Steinbeck,C. et al. (2003) The chemistry development kit (CDK): an open-source Du,Q.S. et al. (2008a) Multiple field three dimensional quantitative java library for chemo- and bioinformatics. J. Chem. Inf. Comput. Sci., 43, structure-activity relationship (MF-3D-QSAR). J. Comput. Chem., 29, 211–219. 493–500. Du,Q.S. et al. (2008b) Recent advances in QSAR and their applications in predict- Stewart,J.P. (1990) MOPAC: a semiempirical molecular orbital program. ing the activities of chemical molecules, peptides and proteins for drug design. J. Comput. Aided Mol. Des., 4,1–103. Curr. Protein Pept. Sci., 9, 248–259. Todeschini,R. et al. (2010) Molecular Descriptors for Chemoinformatics.Wiley-VCH Du,Q.S. et al. (2009) Fragment-based quantitative structure-activity relationship GmbH & Co. KGaA, Weinheim. (FB-QSAR) for fragment-based drug design. J. Comput. Chem., 30, 295–304. van de Waterbeemd,H. et al. (2003) ADMET in silico modelling: towards prediction Gola,J. et al. (2006) ADMET property prediction: the state of the art and current paradise? Nat. Rev. Drug Discov., 2, 192–204. challenges. QSAR Comb. Sci., 25, 1172–1180. Vin˜ a,D. et al. (2009) Alingment-free prediction of a drug-target complex network Gonza´ lez-Dı´az,H. et al. (2005) Predicting multiple drugs side effects with a general based on parameters of drug connectivity and protein sequence of receptors. drug-target interaction thermodynamic Markov model. Bioorg.Med.Chem., 13, Mol. Pharm., 6, 825–835. 1119–1129. Wang,J.M. et al. (2011) Recent advances on aqueous solubility prediction. Comb. Gonza´ lez-Dı´az,H. et al. (2008) Predicting antimicrobial drugs and targets with the Chem. High Throughput Screen., 14, 328–338. MARCH-INSIDE approach. Curr. Top. Med. Chem., 8, 1676–1690. Wegner,J.K. (2005) JOELib: Graph/Data Mining and Clustering. ACS Meeting. Gonza´ lez-Dı´az,H. et al. (2011) MIND-BEST: web server for drugs and target dis- Wei,H. et al. (2009) Investigation into adamantane-based M2 inhibitors with covery; design, synthesis, and assay of MAO-B inhibitors and theoretical ex- FB-QSAR. Med. Chem., 5, 305–317. perimental study of G3PDH protein from Trichomonas gallinae. J. Proteome. Yan,J. et al. (2012) Comparison of quantitative structure-retention relationship Res., 10, 1698–1718. Hanwell,M.D. et al. (2012) Avogadro: an advanced semantic chemical editor, visu- models on four stationary phases with different polarity for a diverse set of alization, and analysis platform. J. Chemoinform., 4,17. flavor compounds. J. Chromatogr. A, 1223, 118–125. Izrailev,S. et al. (2004) Enzyme classification by ligand binding. Proteins, 57, Zhu,J.Y. et al. (2011) Recent developments of in silico predictions of oral bioavail- 711–724. ability. Comb. Chem. High Throughput Screen., 14, 362–374.

Journal

Bioinformatics – Oxford University Press

Published: Mar 15, 2013

There are no references for this article.