Skip to main content
EU Science Hub
News article23 July 20212 min read

Machine learning and ‘omics arrive in the field of regulatory toxicology

The JRC has published an independent scientific peer review of two Genomic Allergen Rapid Detection (GARD) test methods for the assessment of chemicals inducing skin allergies
© European Union, 2021

The JRC has published an independent scientific peer review of two Genomic Allergen Rapid Detection (GARD) test methods for the assessment of chemicals inducing skin allergies.

The JRC’s EU Reference Laboratory for alternatives to animal testing (EURL ECVAM) tasked its Scientific Advisory Committee (ESAC) to assess the scientific validity of two Genomic Allergen Rapid Detection (GARD) test methods. The EASC Opinion is now available and includes in annex the detailed findings of the ESAC Working Group.

The GARD test methods make use of a machine learning algorithm (Support Vector Machine) to process genomic data to identify chemicals that can induce skin allergies (GARDskin) and to distinguish between strong and weak sensitisers (GARDpotency), as described by the Globally Harmonized System for Classification and Labelling of Chemicals.

“This is the first time a machine-learning algorithm has been independently reviewed for use in the field of regulatory toxicology. Using the code, explanations and data provided by the test developer, ESAC was able to verify the algorithm and reproduce the processing steps to convert raw data into predictions for both GARDskin and GARDpotency”, explains Silvia Casati, JRC scientist and coordinator of the review.

ESAC concluded that GARDskin can contribute to skin sensitisation hazard identification in a weight-of-evidence approach. Depending on the regulatory context, positive results obtained with GARDskin may be used in isolation to identify skin sensitisers. However, a negative result obtained with this assay may not be sufficient evidence on its own to identify non-sensitisers and thus should be considered together with other evidence.

ESAC does not consider the information currently available on GARDpotency to be sufficient to recommend its use for regulatory purposes to assess the potency of skin sensitising chemicals. The use of the GARDpotency test to discriminate between Cat 1A and 1B sensitisers is currently not supported by ESAC due to shortfalls in reproducibility and predictive capacity and issues with the design of the validation study.

João Barroso, JRC scientist and ESAC coordinator adds that, “According to ESAC, the GARDskin is ready for further consideration at the OECD for possible development of a Test Guideline. These standard OECD Test Guidelines are internationally recognised test methods used by government agencies, industry and independent laboratories all over the world to determine the safety of chemicals used in many different sectors, including cosmetics, pesticides and industrial chemicals.”

The JRC co-led the recent project at the OECD to develop a ‘Guideline on Defined Approaches for Skin Sensitisation ’ that provides an alternative to conventional animal testing. This Guideline represents the first example of an international standard for chemical safety assessment based on the integration of non-animal data from both experimental and computational methods.

For over 30 years, the JRC has been an international leader in supporting the validation, expert peer-review and international regulatory adoption of non-animal methods that provide mechanistic toxicological information on chemicals to ensure they don’t cause harm to human health or the environment.

As set out in Directive 2010/63/EU on the protection of animals used for scientific purposes, the ultimate goal of EU policy is the full replacement of animal experiments as soon as scientifically possible.

Related Content

ESAC Opinion on the Scientific Validity of the GARDskin and GARDpotency Test Methods

New milestone reached in alternatives to animal testing for chemicals inducing contact allergy


Publication date
23 July 2021