AMBIT Chemoinformatics Software for Data Management Joanna Jaworska

AMBIT Chemoinformatics Software for Data Management Joanna Jaworska P&G Brussels, Belgium Nina Jeliazkova Ideaconsult Ltd. , Bulgaria

Introduction – why Ambit ? Limited free, publicly accessible, methodologically transparent software was identified as one of the roadblocks for broadening use of in-silico methods (ICCA Workshop in Setubal 2002, OECD) Realization that efficient use of existing information on chemicals requires better ways for • Storage − standardized formats, computer automated verification of structures, capability to store large amounts of data • Taking advantage of rapidly evolving field of data mining and extraction of relevant information

IT strategy Ambit - building blocks for Decision Support System High emphasis on • interoperability for “plug and play” • Flexibility modular design • Transparency − Open source, relying on open standards. Open source software lowers the user barrier, facilitates the dissemination activities and enables the reproducibility of models and results − The cheminformatics functionality relies on the open source Java library – The Chemistry Development Kit http: //cdk. sourceforge. net/ − The software is based on My. SQL database (www. mysql. com), which is the most popular open source relational database. − Chemical Markup Language (CML) • acknowledged method of encoding chemical data in XML • Is being adopted by a large number of chemical organisations, from government, through commercial to academia. • The choice of CML for the internal format makes the database independent of the software which is able to access it, in contrast to some proprietary solutions.

Ambit - Overview AMBIT software is a set of libraries and tools, providing various cheminformatics functionalities for data management. The AMBIT system consists of a database and functional modules allowing a variety of flexible searches and mining of the data stored in the database. The unique feature of AMBIT is the ability to store multifaceted information about chemical structures and provide a searchable interface linking these diverse components.

Ambit overview The AMBIT database: • stores chemical structures, their identifiers such as CAS, INCh. I numbers; attributes such as molecular descriptors, experimental data together with test descriptions, and literature references. The database can also store QSAR models. In addition the software can generate a suite of 2 D and 3 D molecular descriptors. • can be searched by identifiers, attribute value or range, experimental data value or range, user defined structure and substructure, structural similarity • AMBIT database contains over 450 000 chemical compounds with data imported from over a dozen databases [http: //ambit. acad. bg/ambit/stats/]. The number of compounds is growing all the time and one the of system’s great strengths is that any dataset can be imported for comparison and analysis. AMBITDatabase. Tools 1. 10 allows the user to create a local database and to import his own sets of chemical compounds. AMBIT Discovery performs chemical grouping and assesses the applicability domain of a QSAR offering a variety of methods including using different approaches to similarity assessments: statistical that rely on ‘descriptor space’; approaches based on mechanistic understanding; and approaches based on structural similarity. Tox. Tree is a flexible user friendly application which integrates structure based (classification) schemes. Currently 3 schemes are available: Verhaaar for fish toxicity, Cramer for human acute toxicity, Bf. R rules for skin irritation. Tox. Tree implements a plug-in mechanism, allowing to be extended by modules developed at a future time, without recompiling the application. Tox. Tree and AMBIT modules can be integrated one within another. Toxmatch – stand alone application for pairwise similarity assessments with intention for read-across. QSAR database under development. Will store information in QMRF. Large effort on standardization

AMBIT Database Today Not restricted to these datasets!Any dataset can be imported. (e. g. DSSTox, AQUIRE, LLNA …)

AMBIT Database Schema

Experimental results repository

Ambit database Two user interfaces to the database • Online • Standalone Online • a more restricted interface Standalone • Full interface • Can be used for storing & managing confidential data Common • Can link with other databases and pull information via webservices

AMBIT database functionalities Storage: information about chemicals name and structure, descriptors, experimental data and QSAR models • Example with a tailored template : BCF golden database LRI project ( EURAS) Q 2 2007 • QSAR database with QMRF ( ECB funded) Conversion: • Different computer formats of structure, CAS-structure Calculation • Variety of descriptors • The available list is growing thanks to contributions to CDK Search • identification search (CAS, SMILES, chemical name) • Descriptor search • Experimental data search • Substructure and similarity search Complex searches with multiple criteria (standalone)

What kind of searches are desired ? • Detailed analyses for pairwise similarity • Similarity of a compound to compounds in the database • Similarity of a compounds to a reference set • Similarity of a set of compounds to compounds in the database • Grouping based on chemical class

Ambit online Searching for basic information

AMBIT Online: Similarity search – replace with new search results !!!

AMBIT Online: Query result

Links to other databases: (example: KEGG)

Link to Aquire

Information about QSAR models

Ambit Database Tools 1. 20 Standalone application available at http: //ambit. acad. bg/downloads

Ambit converter (Batch search) Ambit converter can open : CML, CSV, HIN, ICHI, INCHI, MDL MOL, MDL SDF, MOL 2, PDB, SMI, TXT and XYZ file types Ambit converter can save : SDF, MOL, CSV, TXT, SMI file types. • CAS-SMILES conversion based on a database lookup • Descriptors calculation • Cramer rules, • Verhaar scheme

Ambit Database Tools 1. 20 Import to Database • Compounds – several file formats • Descriptors – SDF, CSV, TXT • Experimental data – SDF, CSV, TXT • QSAR models – SDF, CSV, TXT Database processing • Calculate SMILES/Fingerprints/Atom environments – necessary in order to perform substructure and similarity search. Should be invoked after importing compounds into database • several file formats • Descriptors calculation • Distances calculation – used to speed up distance between heavy atoms query

Ambit Database Tools 1. 20 • perform a CAS RN search in the database (submenu "Search -> CAS RN search"); • perform a SMILES search in the database (submenu "Search -> SMILES"); • perform a molecular formula search in the database (submenu ("Search -> Molecular formula"); • define structure, descriptor, distance-based and experimental data criteria and perform searches in the database • Output: • On screen • To file The user can select between the different datasets existing in the AMBIT database. Subsequent searches will be performed only within the selected dataset

AMBIT User Interface Example: Search by structure • Exact search • Substructure search • Similarity search • Fingerprints • Atom environments

AMBIT User Interface Example: Search by descriptors

AMBIT User Interface Example: Search by experimental data

Similarity based on toxicity mechanism Verhaar scheme Verhaar H. J. M. , Van Leeuven C. , Hermens J. L. M. , Classifying Environmental Pollutants. 1: Structure-Activity Relationships for 34 rules 5 classes Prediction of Aquatic Toxicity, Chemosphere, Vol. 25, No. 4, pp. 471 -491, 1992 • Class 1. Narcosis or baseline toxicity • Class 2 Less inert compounds • Class 3 Unspecific reactivity • Class 4 Compounds and groups of compounds acting by a specific mechanism • Class 5 Not possible to classify according to these rules

Chemical similarity assessment using the database Exact substructure search based on 2 D Structural Similarity search (various methods) Criteria on descriptors Based on mechanistic understanding ( Verhaar scheme)

Another view on Similarity assessments with Toxmatch and Discovery • similarity to a set (summary representation) Toxmatch • pairwise similarities • Similarity to a set (nearest neighbours)

Thank you Questions?