The GARD® technology.

A holistic approach

GARD®, Genomic Allergen Rapid Detection™, is an in vitro test platform developed with a holistic persective. By integrating genomics and machine-learning technologies, the GARD assays are designed to capture the complex biological processes underlying immune responses, such as skin sensitization.

GARD®skin (OECD TG 442E), the assay for skin sensitization hazard assessment, illustrates this approach. Its 196-gene signature is derived from whole-transcriptome profiling, providing an unbiased and comprehensive representation of the cellular and immunological response. Although the gene selection is fully data-driven, many genes within the signature are well known to participate in established Key Events of the skin sensitization Adverse Outcome Pathway (AOP), including:

  • Metabolic activity and identification of pre-/pro- haptens, e.g. ALDH, NAT-1, cytochrome P450 (CYP) enzymes
  • Oxidative stress and cytoprotective responses, e.g. Keap1–Nrf2–ARE pathway (NQO1, HMOX1)
  • Pro-inflammatory signalling, e.g., TNF-α, IL-8, IFN-γ
  • Antigen recognition and innate immune activation, e.g. TLR4, TLR6
  • Dendritic cell migration and maturation, e.g. CD86, MAPK signalling pathways

Figure 1 Genes in the GARDskin biomarker signature mapped to the AOP for skin sensitization

Key elements

All GARD assays are built on a common design principle: the analysis of the comprehensive response of human cells following exposure to test substances.

The GARD technology platform comprises four core elements: a biologically relevant human cell system, a curated training dataset, a genomic biomarker signature, and a machine-learning–based prediction model.

  • Biological cell system

    The biological cell system used by all the GARD tests is a human dendritic-like cell line, SenzaCell®, which mimics a critical part of the human immune system and is able to recognize allergens.

  • Training dataset

    The training dataset consists of genome-wide gene expression profiles from cellular exposures to a set of well-characterized chemical sensitizers and non-sensitizers. The training dataset can be used both for biomarker discovery and subsequent training of the prediction model.

  • Genomic biomarker signature

    By performing gene expression pattern analysis of the training dataset, relevant genes can be selected as the biomarkers for the toxicological endpoint, thereby establishing a so-called genomic biomarker signature.

  • Prediction model

    Based on the gene expression patterns of the genomic biomarker signature studied in the training dataset, machine-learning technology is used to create prediction models for the investigated endpoint. These models are used to perform future classifications of test chemicals.

High performance with human relevance

Traditional in vitro tests investigate only a few biomarkers and provide limited information to give reliable results. Animal tests provide much more information which, however, are not always human-relevant. By using a genomics-based approach with machine-learning technology, GARD combines the simplicity of in vitro methods and the biological intricacy of in vivo models.

This holistic approach contributes to improved accuracy and human relevance. For example, the predictive accuracy of animal tests for skin sensitization assessment is estimated from 70% to 75%  while GARDskin has a predictive accuracy of over 90%.

Broad applicability

GARDskin works for a wide variety of test chemicals, with demonstrated applicability to evaluate “difficult-to-test” samples, including:

  • Complex mixtures
  • Indirectly acting haptens
  • Lipophilic compounds
  • Metal and metal salts
  • Solid materials
READ MORE ON OUR "DIFFICULT-TO-TEST" PAGE