S3: QsarDB Repository

Organization and efficient use of in silico predictive models in chemistry and related areas – QsarDB Repository
physic-chemical properties to human health endpoints, OpenTox Euro 2016
PRESENTING AUTHOR: 

Uko Maran

INSTITUTION / COMPANY : 

Institute of chemistry, University of Tartu, Estonia

POSITION: 

Senior Researcher

ABSTRACT CONTENT / DETAILS: 

The presentation gives an overview of perspective reuse of published models from physic-chemical properties to human health endpoints.  

Presentation also looks into web services, integrated web modelling environments and web repositories available for (Q)SAR models.

The presentation looks more specifically into the QSAR DataBank (QsarDB or in short QDB) approach for the systematic digital organization and archiving of (Q)SAR model information and QsarDB interactive web repository (http://qsardb.org/).

The repository [J. Cheminf. 2015, 7:32] stores models in an open QsarDB data format [J. Cheminf. 2014, 6:25] that is a general chemoinformatic solution for the electronic organization and archiving of (Q)SAR model information according to open standards.

The repository is smart; allows visualization of data and models and has prediction functionality and estimates for applicability domain and provides prediction reporting. For (Q)SAR-s with integrated descriptor calculator web service for predictions is available.

At present moment the QsarDB repository contains over 400 unique models for about 70 physicochemical, toxicological, biochemical and human health endpoints and about 20 species.

Available (Q)SAR-s utilize from simple regression models to complex classification, decision tree, neural network, random forest, support vector machine and consensus model mathematical representations.