Challenge 28

RespiraTox

Launched Awarded Completed

The aim of this Challenge was to develop a QSAR-based tool that reliably predicts the human respiratory irritancy potential of chemicals. The tool should fulfill the five OECD principles for QSAR validation to demonstrate the statistical and mechanistic reliability of the model. This will endorse the model's use under regulatory context (e.g. REACH, Environmental Protection Agency (EPA)).

To address this Challenge, Dr Sylvia Escher (Fraunhofer ITEM, Germany) and Dr Andreas Karwath (University of Birmingham) have developed a QSAR model that is available as a web-based tool, which allows end-users to predict human respiratory irritation of chemical compounds by entering structural information.

Challenge completed

Through the RespiraTox Challenge, Dr Sylvia Escher (Fraunhofer ITEM, Germany) and Dr Andreas Karwath (University of Birmingham) have developed a QSAR model that is available as a web-based tool, allowing end-users to predict human respiratory irritation of chemical compounds by entering structural information.

Conference presentation

Society of Toxicology 58th Annual Meeting (Baltimore, USA)

Development of a QSAR Model to Predict Respiratory Irritation by Individual Constituents.

Challenge awarded

A team led by Dr Sylvia Escher from the Fraunhofer Institute for Toxicology and Experimental Medicine has been awarded £99,996 to deliver the project: RespiraTox: In silico model for predicting human respiratory irritation.

Challenge launched

Sponsored by Shell, the RespiraTox Challenge aims to develop a QSAR-based tool that reliably predicts human respiratory irritancy potential of chemicals. The tool should fulfil the five OECD principles for QSAR validation to demonstrate the statistical and mechanistic reliability of the model. This will endorse the model's use under regulatory context (e.g. REACH, Environmental Protection Agency (EPA)).

Background

Inhalation of certain chemicals may potentially cause irritation to the respiratory tract resulting in inflammation, which if unresolved can lead to irreversible fibrosis of the lungs (Cometto-Muñiz and Cain, 1995). Examples of respiratory irritants include acetic acid, benzoyl chloride and formic acid.

Currently there are limited in silico, in vitro and in vivo models to determine the respiratory irritation potential of new or existing substances. Assessing whether a chemical will cause respiratory irritation in humans is often determined by observations in rodent acute (single) and repeat dose inhalation toxicology studies. However, there are no specific test protocols in place to determine the irritancy potential of respiratory toxicants or allergens. In the absence of specific or well defined guidelines, respiratory irritation results are extrapolated from the acute inhalation toxicity studies (the Organisation for Economic Co-operation and Development (OECD) test guidelines (TG) 403 and 436) performed on rats (OECD, 2009a, OECD 2009b). This involves the modification of protocols to include endpoints for respiratory irritation and requires additional dose groups. For example, RD50 data (concentration producing a 50% respiratory rate decrease as determined by the Alarie test (Alarie Y, 1966)) in rodents is often used as a surrogate for irritation potency of respiratory irritants. However, it is difficult to extrapolate the rodent respiratory hazard data to human respiratory irritation.

Under the REACH (Registration, Evaluation, Authorisation and Restriction of chemicals) regulations, the registrant may be able to demonstrate that a substance poses no respiratory risk if exposure via the inhalation route is not expected. However, for most substances exposure via the inhalation route is likely to be common, and if the substance is a skin or eye irritant then it may be difficult to justify a waiver for acute inhalation studies. Without robust models for respiratory irritation, it is possible that chemicals may pass through the R&D pipeline and reach the market place with the potential liability of being respiratory irritants. The goal of this Challenge is to develop an accurate in silico tool that is capable of predicting human respiratory irritation potential.

3Rs benefits

The respiratory irritancy potential of chemicals is typically assessed and extrapolated from modified rodent acute inhalation toxicity studies (OECD TGs 403 and 436 (OECD, 2009a, OECD, 2009b)). These in vivo toxicity studies are classified as severe under the UK’s Animals (Scientific Procedures) Act and require additional dose groups. A typical modified acute inhalation toxicity study uses approximately 42 animals/study.

Development of a QSAR tool that reliably predicts the respiratory irritancy potential of chemicals in humans will allow for the early identification of potential toxicities in candidate chemicals without having to use in vivo studies, and contribute to the scientific justification to waive the in vivo studies for respiratory irritation for those taken forward to registration.

Single Phase Challenge winner

Project team led by:

Full Challenge information

PDF - 52.36 KB

PDF - 156.88 KB

PDF - 247.03 KB

Assesment information

Review and Challenge Panel membership

 

Dr Sylvia Escher, Matthias Wehr (Fraunhofer ITEM, Germany) and Dr Andreas Karwath (University of Birmingham, UK), in collaboration with Challenge Sponsors Shell, have developed a QSAR model to predict human respiratory irritation, allowing the early identification of potential toxicities in candidate chemicals without having to use in vivo studies.

The model was trained using 2153 organic compounds and uses binary data on the absence/occurrence of respiratory irritation differentiating sensory irritation from histopathological effects including tissue irritation or damage. These data were curated from several sources*.

In a 5-fold cross validation of the training data, the model’s performance was measured using the area under the receiver operating characteristic (ROC) curve (AUC) and the current model achieved an AUC of 0.71. The model shows high sensitivity, but relatively low specificity meaning that compounds may be falsely predicted as being an irritant to the respiratory tract. The model could therefore be used for prioritisation and screening.

The combined model including sensory and tissue irritation is available as an open access web-based tool, allowing end-users to predict human respiratory irritancy of organic chemical compounds by solely entering structural information. The tool provides a readout of the confidence of the prediction as given by the underlying Random Forest algorithm and whether the structure is within the models applicability domain of the training dataset. Finally, a list of structurally similar analogues are given, and the end-user is able to choose between three different models to calculate similarity. Furthermore, for closely related analogues, the available data on respiratory irritation from the different curated original sources are displayed for further investigation. The overall approach adheres to the five OECD principles for QSAR validation.

The user-friendly tool can be accessed here.

Further information about the project can be found here.

* List of data sources

Human respiratory irritation model

A QSAR model to predict human respiratory irritation of organic compounds

A team from Fraunhofer ITEM (Germany) and the University of Birmingham (UK) has developed a QSAR model to predict whether candidate chemicals cause human respiratory irritation.

The model is available as an open access web-based tool, allowing end-users to predict human respiratory irritancy of organic chemical compounds by entering structural information. The tool provides a readout on the reliability of the prediction and whether the structure is within the models applicability domain of the training dataset.

The user will be able to control the prediction by comparing it to a list of structurally similar neighbours, for which the available data on irritation is displayed ready for further investigation.

The user-friendly tool can be accessed here.