The Marvin family MarvinSketch MarvinView MarvinSpace Available as Java applets for HTML pages and Java beans for standalone apps (full API) Structure, query & reaction editing Individual and structure table visualization Publication quality macromolecule visualization MarvinSketch/View http:// www.chemaxon.com/MarvinSketch_View.ppt MarvinSpace http:// www.chemaxon.com/MarvinSpace.ppt
Marvin D evelopment Histor y 1998 Applets, Molfiles, stereo support, Windows, Unix SMILES , SMARTS, PDB, Rgroups , isotopes, shortcuts, Marvin Beans Ball and stick JPG, PNG, SVG, Cut&Paste with Isis/ChemDraw, 2D cleaning , (de)aromatization, reaction drawing 2000 1999 SDF, RDF, XYZ animations, CML, templates, compressed formats, Swing, 3D models 2001 Mac support, signed applets, Java Web Start, atom mapping Partial charge, p K a , log P / log D , 3D optimi z ation , radicals , abbreviated groups Marvin file format, enhanced stereo , shapes, text boxes, multiple groups, link nodes, TPSA, recursive SMARTS , D onor/Acceptor , electron arrows, 200 4 2003 200 5 Tautomers , resonance , lone pairs , conformers, 3D sketching , MarvinSpace , Topology ana l y s i s , presentation quality graphics , .. . 2006 More Plugins, more R-groups, EMF, PDF and Mol2, Improved property storage in MRV, SDfiles and Rdfiles. .NET support in MarvinBeans. 2002 2007 Structure to name , Coordinati on compounds, Polymer drawing, OLE , Markush enumeration plugin Configurations 2008 Name to structure , OLE 2 , Chemical Terms Customizable GUI
Calculator Plugins Calculator Plugins h ttp://www.chemaxon.com/Calculator_Plugins.ppt
Standard IUPAC Name
pKa, Major Microspecies, Isoelectric Point
Charge, Polarizability, Orbital Electronegativity
Tautomerization, Resonance, Stereoisomer
Conformer, Molecular Dynamics
Topology Analysis, Geometry, Polar Surface Area (2D), Molecular Surface Area (3D)
Hydrogen Bond Donor-Acceptor, Huckel Analysis, Refractivity
A vari e ty of structure based calculations are available from the Marvin GUI, cxcalc command line tool and the API. The calculations are widely used within several JChem tools and are available as functions of Chemical Terms expressions.
Chains, Monocycles/ Traditional names with and without heteroatom/ Spiro ring systems/ Ethers/ Common characteristic groups, Ionic compounds/ Unlimited number of atoms and rings/ All atom types /Stereochemistry/ etc.
drag&drop or copy&paste to MarvinSketch
Label updated in real-time
Automatic format recognition
Batch from command line
JChem family JChem Base JChem Cartridge Instant JChem Fast substructure and similarity searching Tight Oracle SQL integration Desktop application for scientist ChemAxon’s proprietary database structure Arbitrary database structure Access local and remote databases
JChem development history 2000 Oracle , MySQL, SQLServer, Access, hashed fingerprints , substructure and similarity search DB2, PostgreSQL, Rgroup searching R eaction searching , fragmentation, r eact ion processing , standardization, pharmacophore s , screening 2002 2001 Clustering , diversity 2003 R-decomposition , R- enumeration , reaction library , custom fingerprints, random s y nthesis , link nodes … 2005 2006 T automer search , Instant JChem reaction similarity, Library MCS , GUI for Standardizer/ Reactor … 2007 Calculated columns, Installer, Tautomer Duplicate filtering, Query tables, Markush tables , Speed enhancements for JChem Cartridge , form design, relational data for Instant JChem ... C artridge , enhanced stereo searching, recursive SMARTS , Chemical Terms , virtual s y nthesis 2004 2008
The learning curve of chemists familiar with ISIS is very short. After having some practice, Marvin is reported a more productive drawing environment. The most of the MDL features are available in Marvin and JChem, and many others not available in MDL technology.
FMC migrated from MDL® ISIS/Base ISIS/Host to ChemAxon’s JChem. They later published their detailed scientific comparison.
Used 1.8 million vendor compounds to create a testing database
Prepared 115 different query structures for comparison
51 simple sub-structure search
51 similarity search
64 complex search
Identical search hits in almost all cases, major differences result from MDL’s incorrect aromatic bond definitions in case of 5 member aromatic rings. ChemAxon's approach is the chemically correct and their performance is higher (faster).
1. What was the platform you used before the migration?
All systems were run using the Daycart cartridge on Linux servers
MDL Cartridge running on Sun Solaris
We used ISIS/Host as a server, the client was ISIS/Base customized using ISIS/PL
Daylight and IDBS Chembridge
2. How long did it take to migrate?
Very simple, hardly any time at all, just a few hours to uninstall old cartridge, install new cartridge and build indexes. Then modify a few SQL statements in the code to use the new cartridge functions.
It took a full weekend to switch over and convert all old databases.
Since we use SQL for structure searches, the actually change in the application code are few. Code changes takes about 1 day. However, we spent at least two weeks to compare the daylight and jcart.
It took 1 year for planning, and another 1 year for designing and developing the system. 1-year-migration time includes all of the operation that is needed. That means our technical people worked for this project 1 year. We migrated the data structure of HView, but the form was re-designed in order to fit our existing (wet) workflow.
3. How many technical people were required in the migration process?
It was fairly simple so just one developer with all round programming, database, and chemistry knowledge.
6 technical people. 2 were contacting with users. For the system design, 11 users were involved from chemistry, HTS, eADME groups.
4. Why did you decide on leaving the previous platform? (problems)
Purely the cost. We found the Daycart system to be very good, very stable, fast, and the API was well thought out. However, it was just too expensive for us.
Old technology not offering new functionality. High cost, in particular for new licenses.
Daycart (at least at that time) did not take MOL query, not all query structures could be correctly presented as smiles/smarts.
Two main reasons were the maintainance cost, and the accessibility. We had to suppress the raising system (software) cost, and at the same time we had to enlarge the number of users and client PCs from which we could use DB system.
5. What alternative platforms were considered/evaluated?
Prior to selecting ChemAxon we looked at all the cartridges available at the time
The Accord cartridge was also evaluated. Some others did not qualify for evaluation.
Accord (Accelrys), and ChemOffice (Cambridge Soft) were two major alternatives.
Symyx/MDL Direct Oracle cartridge
6. Why did you choose ChemAxon technology? (advantages)
Cost was a major factor, but also because we felt we could work with ChemAxon to develop the tools further as we wanted to use them. A very open approach. Another reason was that all the tools we needed were available from a single vendor, i.e. Oracle cartridge for searching, and sketching and viewing tools.
Almost as good as Accord but with better impact on improvement and support.
Marvin Sketch and JCart represent the molecules in MOL using exactly the same backend library. MOL is used instead of smiles/smarts. Much faster search. Price is good .
We could keep the cost lowest by using ChemAxon, and more than that, the affinity for the web technology was favorable to our future vision of the cheminformatics system.
The greatest advantage is the low cost and great support. We have always had MDL/Direct cartridge, but the greatest advantage is the low cost and stellar support speaks specifically to ChemAxon.
7. What were the most problematic issues occurred during the migration? (negative impressions)
Understanding the finer points of all the search functions / options i.e. precisely how things like aromaticity, stereochemistry, etc. are handled. We've also had to spend time considering how to restandardise structures and how to rewrite SQL. When doing a straight forward structure search (i.e. benchmarking), the JChem cartridge performs very well against other systems such as Daylight, however, if you want to incorporate joins between tables can considerably affect the query times even when using what we call ChemAxon SQL.
Structure matching bugs in the cartridge and undocumented actions needed to be performed.
JCart installation was not so smooth 3 years ago. Much better now. Most of the problem and issues are because some structures are interpreted different between the two software. Some are Daylight bugs and others are jchem bugs. JChem has fix all their share.
There were little problem, what I remember is that the response was slower than expected when the chemical object was included in the page.
8. How could you overcome in these difficulties? (resolutions)
We spent a lot of time experimenting with the different functions/options so we completely understand what they do.
The structure search bugs was overcome by rewriting the registration procedures, undocumented actions were overcome by hard work.
Wait until major bugs in JChem are fixed. We live with about 0.01% of inconsistencies and work it out later.
The needless chemical objects were replaced by pictures.
Availability and quick turn around to patch any
9. Did you expect any other problem, that did not occur? (positive impressions)
We though there may be problems running two different cartridges on the same table but this worked fine
Not really. Most MDL features were available in JChem. This was one of the selection criteria, particularly important for chemical registration.
We expected that the transfer of the existing data might be problematic, and that the system change might be inconsistent with existing 'wet' workflow. That was why we organized 11 users as a system designing team, and I think the team worked well.
10. What additional components were purchased together with the JChem Cartridge?
Most of them!
None. User probably should consider plug-ins for calculating HBD, HBA, logp, psa, etc. We did not because we need to stick to CLOGP in order to be consistent with the rest of the company.
11. How much technical support did you need from ChemAxon for the migration?
Initially quite a lot, though the products have been developed a lot since then. We haven't required much support for structure migration, but we've also migrated a load of SMIRKS and we've needed support for that mainly because of the way in which they were handled in the old system (non-standard).
A few needed support cases where filed on the support forum and fairly quickly resolved.
Lots, we had close communication with dev team during the migration.
Our technical people sent e-mail several times to your support team.
14. Are you satisfied with the performance/functions of the ChemAxon powered system?
The number of functions available and flexibility of the JChem tools is excellent, and allows us to develop very interesting and useful drug discovery software for our scientists.
Useful migration resources ChemAxon's Marvin & JChem (v 3.1.3) vs. MDL® ISIS/Draw ISIS/Host (v 4.0) Seong Jae Yu, David Roush*, Usha Ganesh, Young Moon, Henry Liu, FMC Corp. http://www.chemaxon.com/conf/FMC_ChemAxon_JCHEM_Cart_xnotes.ppt User Group Meeting presentations: http://www.chemaxon.com/UGM/ugm_land.html
“ Cheshire is a scripting language that enables you to write scripts to validate, modify, or gather information about chemical structures, such as molecules and reactions.”
What alternatives can ChemAxon offer?
ChemAxon’s Java API (also available from .NET)
Java API for Cheminformatics from ChemAxon ChemAxon’s class library consists of more than 1500 chemistry related classes tuned for usability and high performance.
Chemical Terms charge () and match ( amine ) or match ( hydrazine ) Chemical Terms offers more than a hundred popular chemistry functions opening up the power of cheminformatics for those scientists who focus on quick results instead of the details of programming and scripting. The integration of Chemical Terms makes make chemistry applications smarter and more customizable.
Standardizer for Batch Conversion Standardizer is a batch conversion utility providing many useful and customizable functions for the canonicalization of chemical structures and restoration renovation chemical information in structures from older databases.
Standardizer Actions Aromatize Dearomatize Add Explicit Hydrogens Remove Explicit Hydrogens Clean2D Clean3D Transform Wedge Clean Clear Isotopes Remove Fragments Remove R-groups Neutralize Tautomerize Mesomerize Set Absolute Stereo Remove Absolute Stereo Convert Wedge Interpretation Convert Double Bonds Clear Stereo Alias to Group, Alias to Atom Contract Group Expand Group Ungroup Expand Stoichiometry Map Reaction Unmap
Counting Groups – Cheshire Counting O=S=O groups in Cheshire
Counting Groups – Java API Counting any functional groups with ChemAxon’s Java API Counting O=S=O groups in Chemical Terms
Adding Explicit Hydrogens - Cheshire Adding explicit hydrogens and cleaning the molecule in Cheshire
Adding Explicit Hydrogens – Java API Adding explicit hydrogens and cleaning the molecule with ChemAxon’s Java API
Adding Explicit Hydrogens – Standardizer Adding explicit hydrogens and cleaning the molecule with Standardizer The same in command line
Group Conversions – Cheshire Conversion of neutral form of nitro to the ionic one in Cheshire
Group Conversions – Java API Conversion of neutral form of nitro to the ionic one with ChemAxon’s Java API
Group Conversions – Standardizer The same in command line Conversion of neutral form of nitro to the ionic one in Standardizer
ChemAxon’s Java API provides similar freedom and flexibility to Cheshire for programmers to develop chemistry functions for any tears like web clients, desktop applications, server systems and Oracle stored procedures.
Java is a standard language with worlwide community, rich resources and lots of well educated developers. (The ChemAxon Java API is also accessible from .NET.)
Chemical Terms provides more than a hundred high level, ready to use functions substituting dozens of lines of complex Cheshire code.
Chemical Terms expressions can directly be used in database filters, virtual reactions, pharmacophore definitions or other cheminformatics applications.
Standardizer is an easy to use batch tool and graphical interface for chemists to create conversion rules without writing a single line of code.
The upcoming Structure Checker will provide and extensible set of quick “problem detection” functions that can be integrated in any applications and will be added to Marvin and Standardizer as well.
Feature comparison to ISIS/Base Feature ISIS/Base Instant JChem Databases Local + Oracle (differences in steroechemistry and calculations etc.). Local + Oracle + MySQL (no differences in local and remote db functionality). Forms Form builder. Form builder. Tabular view Limited. Comprehensive. Relational data Hview DataTree. Deployment Installer only. Installer or Java Web Start. Many deployment features. Collaboration Limited. Many collaboration and sharing features. Scalability & performance Limited, especially for local DBs. Good.
Allows easy extension by ChemAxon, customers and 3 rd parties
Strong enforcement of APIs
IJC functionality is built upon these APIs
Current architecture Local DB IJC Client Database Remote DB Oracle cartridge
IJC server architecture IJC IJC Client Database IJC Server Services API IJC server due Q1 2009 Oracle cartridge Web Apps Web services
Migration issues: general Database artefacts IJC is currently table based. Access to views, synonyms etc. is currently being added. Use of database links has not been investigated yet, but no particular problems expected . Security model
Current implementation provides basic access control:
Edit database model
Users can create forms/lists etc. even in read-only mode, but can’t modify data that affects other users.
Security integration LDAP is probably the most suitable, but the security implementation is quite flexible and customisable.
Migration issues: migration of ISIS DBs Hview vs. Data Tree No direct conversion, so this would currently need to be done manually, though some automation is potentially possible. Data Tree is modelled on the same approach as Hview, so migration in most cases should be relatively simple. Forms No direct import from ISIS, but creating IJC forms is very simple and fast. Customisation Currently no equivalent to ISIS/PL. ISIS applications with complex logic may be more suited as IJC extensions, or as standalone web or JWS applications.
Hview vs. Data Tree: standard tables master_table master_table_id col1 col2 col3 detail_table detail_table_id master_table_id cola colb colc * ISIS Hview HVIEW my_data TREE master DEVICE oracle USERNAME scott PASSWORD tiger TNAME master_table TREE detail DEVICE oracle USERNAME scott PASSWORD tiger TNAME detail_table LINK master (master_table_id) over detail (master_table_id) One-to-many relationship IJC Data Tree
Hview vs. Data Tree: Mol + Rxn tables < RC tables> inventory inventory_id molregno cola colb colc * ISIS Hview HVIEW cpds_inv TREE compounds DEVICE chemicaldb USERNAME CPD/CPD PASSWORD TNAME compounds TREE inventory DEVICE oracle USERNAME scott PASSWORD tiger TNAME inventory LINK compounds (molregno) over inventory (molregno) IJC Data Tree compounds molregno structure [jc_index] inventory inventory_id molregno cola colb colc *
Migration options Simple local or Host based databases used primarily for searching/reporting Migrate to IJC ISIS/Base application with complex application logic but standard (Hview based) data structure (e.g. registration applications) Create custom IJC extension module built upon IJC API (much of the existing IJC functionality is essentially done this way) Application with complex data model and logic Either: Create standalone web application or: Create standalone JWS application or: Create custom IJC module that defines its own data access API.