Our research activities cover three areas:
Multi-scale modelling of toxicity pathways. We develop mechanistic or statistic modelling tools for parameterising Adverse Outcome Pathways for use in a regulatory context.
Understanding and predicting effects of chemical compounds at a population level. We develop mechanistic mathematical models to analyse data obtained from experimental ecosystems in mesocosm studies and to extrapolate from organism-level laboratory data to population viability. This completes the multi-scale modelling of toxicity pathways in ecotoxicology.
Characterisation of individual exposomes with biomarkers. We work on how methodologies for characterisation of exposure can be improved. Integrating exposure dynamics in epidemiological studies can in turn provide better understanding of the link between exposure and health effects.
The field of toxicology is leading to an ever more detailed knowledge of toxicity pathways. Better understanding of biological mechanisms and their stepwise progression towards a pathology will help predict changes due to exposure to chemicals. Danger could then be characterised based on in vitro or in silico studies of these biological pathways instead of by identifying adverse effects in animals at high dose levels. These toxicity pathways, coined Adverse Outcome Pathways (AOPs), are undergoing standardisation at an international level in particular by the OECD, EFSA, and US EPA. AOPs describe the events and the links between events that can lead to an adverse effect following an interaction between a chemical and one or several molecular targets. The AOP concept unifies in vivo, in vitro, and in silico toxicological information obtained at several biological levels of organisation (subcellular, cellular, tissue, etc.). The US EPA ToxCast program is one of the best-known initiatives aiming to support this approach. This program has produced High Throughput Screening (HTS) data on hundreds of key events such as ligand-binding. We contribute to projects which integrate data obtained with advanced in vitro and in silico methods into an AOP framework. To date most AOPs only provide a qualitative description of the links between events, but quantitative inputs can be used to estimate the strength of the relationships between the events. This will allow an estimation of concentration levels that produce biologically significant effects (Points Of Departure).
The computational methods we develop, in particular Quantitative Structure-Activity Relationships (QSAR), physiologically-based toxicokinetic models (PBPK), and systems biology, all fit perfectly in a quantitative AOP framework. They can predict the exposure of living organisms to chemicals, the occurrence of key events, and the mechanistic links between these events. These methods can also integrate information from several biological levels of organisation, which is crucial in AOPs.
We are involved in the EU-ToxRisk, EuroMix, and StemBANCC projects, which all deal with this field of research.
Growing evidence suggests that illnesses, in particular chronic diseases, are due to a combination of genetic and environmental factors. Exposure to environmental factors is extremely variable and complex, and the proportion of chronic diseases attributable to them is largely unknown. The exposome encompasses all environmental exposures from conception onwards. This includes all non-genetic factors: chemical, physical, and biological exposures as well as the psycho-social context and internal biological factors. The exposome was first proposed by Dr. C. Wild in 2005 who noticed that knowledge of genes was far more precise than that of the environmental exposure. Contrary to the genome, the exposome varies throughout a person’s lifetime. Mapping an individual’s exposure is therefore a monumental task. Specific exposures can go unnoticed either because they are not even suspected or due to lack of sufficiently sensitive analytical methods. Exposure indices may be transitory and several specific lifetime periods are now regarded as windows of susceptibility.
Physiologically-based toxicokinetic (PBPK) modelling is one way of characterising individual exposomes. This methodology offers understanding of the relationship between internal concentration levels, measured in biological matrices such as urine or blood, and external concentration levels, measured either directly with sensors or estimated with multimedia models. In the past, we have used a reverse dosimetry approach to reconstruct populations’ external exposure based on measured biomarkers, survey questionnaires, and a toxicokinetic model of the fate of the chemical in the human body. Depending on persistence of chemicals in the organism, exposure biomarkers, usually blood or urine concentrations of parent compounds or metabolites, can represent more or less recent exposures.
The PARC project (former projects: HELIX , HBM4EU, NEUROPHYTO) cover this field of research.
Mesocosms have been proposed as a tool to study the effects of chemicals at the population level. However, studies conducted in mesocosms are limited to a low number of replicates. This, together with the high variability characterising the population level, constitutes a limiting factor for detecting significant effects.
The uncertainty on the distribution of the model variables cannot be narrowed by further increasing the number of replicates in mesocosm experiments, but prior knowledge can be integrated instead. A mathematical model of population dynamics can be used to predict the distribution of the observations under control conditions (Beaudouin et al., 2008 et 2012b). We integrate all available data for our particular experimental setup, recorded in previous years for example, in our model to obtain robust estimates of parameters related to the species under study and to the experimental setup. Better knowledge of the ecosystem functioning under control conditions will allow us to detect disturbances with more precision. We have already applied this method in other experimental systems (Beaudouin et al. 2008, Beaudouin et al. 2012).
The DIADeM project covers this field of research
Endocrine disruptor compounds (EDCs) are of great ecotoxicological concern because of their potential harmful effects on humans and wildlife, including fish. The extrapolation of subtle functional deficits within individuals into population-level effects is a great challenge for the risk assessment of EDCs. Ecological risk assessment should protect the long-term persistence of populations of species in space and time under naturally varying field conditions and in the presence of other stressors (e.g. food limitation). However, excepting the ecotoxicological data provided by mesocosm experiments and a few field studies, data on impacts of chemical substances on populations or higher biological levels are very sparse. In this context, models can play an important role in bridging the gap between what is measured (organism-level endpoints) and what needs to be protected (population-level endpoints).
Mechanistic models: PBPK, Individual-based model, DEB/DEB-tox, systems biology.
Statistical models: QSAR; Bayesian statistics.