Systems ecotoxicology

Our methods and custom pipelines can be applied to aid the understanding of underlying biological changes that occur in a given project.

We utilise a number of computational approaches to understand how stressors such as chemicals, climate change, nutrient deficiency, or predators affect the molecular state of a large variety of organisms. Central to this is the ability to integrate data across different levels of biological organisation and apply statistical/predictive modelling techniques to develop adverse outcome pathways (AOPs), see the figure below.

Adverse outcome pathway

AOPs are a conceptual framework that define the relationships between molecular components central to the interaction with a given stress, called the molecular initiating event (MIE), to the adverse outcome (AO), which is defined as any unwanted adversity, via key events (KEs). KEs can be observed at any level of biological organisation but are required in the progression from the MIE to AO.

These type of analysis pipelines are not limited to questions surrounding environmental health but are also increasingly applied in the biomedical area to conceptualise the knowledge developed during research projects.

Systems (Eco)-toxicology approaches

The CBF continues to strive to improve chemical risk assessment across industry, government, and academia by continuing to develop and pioneer Systems (Eco)-toxicology approaches.

Both Professor Francesco Falciani and Dr Philipp Antczak initially pioneered aspects of the general framework of Systems Ecotoxicology to address limitations regarding the identification of risks presented by the large amount of stress humans impose on the environment every year.

By developing predictive models and statistical linkages between chemical structures, molecular responses, and phenotypic outcomes they have showed that Omics data can be used to improve risk assessment and answer questions about the underlying biological mechanisms that lead to changes in phenotypic response.

Our framework also allows the identification of a chemical toxicity pathway from large scale functional genomics datasets when no other knowledge on a chemical MOA is available. Our group has been able to showcase the use of this framework in several published and unpublished outcomes.

This research has attracted the attention of several UK and international governmental organizations (OECD, the EC, DEFRA) as well as industry (Unilever, AstraZeneca, Thames Water) and has led to the development of several official documents stating the importance of a Systems based risk assessment approach. This involves an OECD AOP for narcosis that we co-developed and it is currently an integral part of the recommended risk assessment procedure for this very important class of chemicals. Changes to various OECD test-guidelines used across the world, and a more Omics based risk assessment procedure are being developed at the US-EPA/EU agencies and Unilever.

Examples of our work

Systems biology approach

In Antczak et al., 2015, we combined chemoinformatics approaches with phenotypic readouts and transcriptomics data to investigate the molecular effect of baseline toxicity. This is a mechanism common to all chemical agents and described as the general relationship between toxicity and lipophilicity of a given compound. 

Linking metabolite profiles of mussels to predict contaminants effects in sexual differentiation and development

In this recently submitted publication, we linked the metabolic state of the blue mussel Mytilus edulis with gonadal development and environmental factors such as temperature. By modelling these variables we created a tool able to infer sex-affecting pollutants in the environment. (Kronberg, et al., 2019).

Analysis pipeline examples

  1. Physico-chemical features (PCFs -chemical structural descrip­tors) were calculated for each com­pound used in our study
  2. The molecular data complexity was reduced by summarizing biological data into pathway activity index (PAI). A PAI is based on principal component analyses and utilises its variance summarisation ability to reduce a pathway such as 'TGF-beta signalling pathway' from 84 genes to 2 PAIs
  3. PAIs were then explained as a function of PCFs which resulted in a list of pathways whose activity could be predicted by a group of structural features. Then PAIs able to predict the observed toxicity were identified
  4. A network modelling framework was used to integrate the results and identify the relationship between structural features (PCFs), pathways (via PAIs) and observed toxicity
  5. Ca2+signalling was identified as one of strongest links with baseline toxicity
  6. Findings were validated showing that close to 43% of the response indicated by a Ca2+ signalling perturbation would recapitulate the baseline toxicity mechanism.

Network modelling

Network modelling example

Using time dependent models to link data over time and infer directionality in networks to better understand the underlying molecular response. Basili et al., 2018.


Biomarker identification

Diagram showing clustering based on differential gene expression

Optimising a small subset of features to build predictive models that can distinguish multiple compound classes. Antczak et al., 2013.


Pipeline development

 Deriving new pipelines for assessing the change in molecular state in complex mixtures

Deriving new pipelines for assessing the change in molecular state in complex mixtures.


Statistical modelling

Statistical modelling example using non-linear modelling techniques

Using non-linear modelling techniques to link multiple levels of data to identify driving forces of changes in biodiversity.

Back to: Computational Biology Facility