Session 5: Data-integration

Data-integration for Endpoints, Chemoinformatics and Omics (DECO) using DIAMONDS
Dinant Kroese, Ph.D: Data-integration for Endpoints, Chemoinformatics and Omics (DECO) using DIAMONDS

Dinant Kroese, Ph.D, Senior Scientist


Risk Analysis for Products In Development (RAPID)


ED Kroese, EP van Someren, J Venhorst, HE Buist, S Bosgra, JTWE Vogels and RH Stierum - Risk Analysis
for Products In Development (RAPID), TNO, Zeist, the Netherlands;

H Kamp, G Montoya-Parra - Experimental Toxicology
and Ecology, BASF SE, Ludwigshafen, Germany;

G Patlewicz - DuPont Haskell Global Centers for Human Health and Environmental Sciences, Newark, DE, USA;
J Polman4, D Jennen - Department of Toxicogenomics, Maastricht University, the Netherlands


Innovations in many industrial sectors involve the development of new chemical entities with improved properties. Though the regulatory frameworks in place may differ for each industrial sector, all strive to prevent harmful effects to exposed humans. Safe limit values are often driven by complex regulatory endpoints such as repeated dose toxicity, carcinogenicity, or reproductive toxicity. Until recently, the hazards of these complex endpoints could only be identified and quantified on a per compound basis by in vivo animal studies.

Today, developments in toxicological sciences, systems biology and computational chemistry, and their integration provide opportunities to predict toxicological profiles of chemicals with highly reduced in vivo testing. At TNO, we are developing DIAMONDS (Data Infrastructure for Applying Models ON Design and Safety), a data infrastructure with statistical and computational tools aimed at predicting complex toxicological endpoints through integrated analysis.

In addition, DIAMONDS includes in vitro effect-specific screening models for ‘biological verification’ for verification of in silico-based toxicity predictions, thus reducing the uncertainty often associated with in silico models for complex endpoints. Part of DIAMONDS was developed in the Cefic-LRI AIMT3 DECO project.

A transparent framework was created for improving prediction of repeated dose toxicity by integrating chemoinformatic data with biological information from ‘omics’ and high-throughput screening (HTS) technologies. The investigated prediction approaches consisted of unsupervised clustering approaches for grouping chemical analogues and supervised class prediction approaches to develop classifiers for different liver toxicity endpoints. Results showed that by integrating different data types the clustering of analogues could be improved.

Furthermore, the conducted classification approaches showed that using omics data leads to a better prediction of repeated dose toxicity. In the FP7 project ChemScreen we explored a battery approach to identify reproductive and developmental toxicants, but also to biologically verify grouping and read across of structurally-related chemicals.

For a good prediction of toxic potency, quantitative in vitro in vivo extrapolation (QIVIVE) proved to be indispensable and is presently being incorporated into DIAMONDS.