Proactive identification of chemicals with skin sensitizing properties is a key toxicological endpoint within chemical safety assessment, as required by legislation for registration of chemicals. In order to meet demands of increased animal welfare and facilitate increased testing efficiency also in nonregulatory settings, considerable efforts have been made to develop nonanimal approaches to replace current animal testing. Genomic Allergen Rapid Detection (GARD™) is a state-of-the-art technology platform, the most advanced application of which is the assay for assessment of skin sensitizing chemicals, GARD™skin. The methodology is based on a dendritic cell (DC)-like cell line, thus mimicking the mechanistic events leading to initiation and modulation of downstream immunological responses. Induced transcriptional changes are measured following exposure to test chemicals, providing a detailed evaluation of cell activation. These changes are associated with the immunological decision-making role of DCs in vivo and include among other phenotypic modifications, up-regulation of co-stimulatory molecules, induction of cellular and oxidative stress pathways and xenobiotic responses, and provide a holistic readout of substance-induced DC activation. Here, results from an inter-laboratory ring trial of GARD™skin, conducted in compliance with OECD guidance documents and comprising a blinded chemical test set of 28 chemicals, are summarized. The assay was found to be transferable to naïve laboratories, with an inter-laboratory reproducibility of 92.0%. The within-laboratory reproducibility ranged between 82.1% and 88.9%, whereas the cumulative predictive accuracy across the 3 laboratories was 93.8%. It was concluded that GARD™skin is a robust and reliable method for the identification of skin sensitizing chemicals and suitable for stand-alone use or as a constituent of integrated testing. These data form the basis for the regulatory validation of GARD™skin.

Toxicological assessments of skin sensitizers have progressed towards a higher reliance on non-animal methods. Current technological trends aim to extend the utility of non-animal methods to accurately characterize skin-sensitizing potency. The GARDskin Dose–Response assay has previously been described; it was shown that its main readout, cDV₀ concentration, is associated with skin-sensitizing potency. The ability to predict potency from cDV₀ in the form of NESILs derived from LLNAs or human NOELs was evaluated. The assessment of a dataset of 30 chemicals showed that the cDV₀ values still correlated strongly and significantly with both LLNA EC3 and human NOEL values (ρ = 0.645–0.787 [p < 1 × 10⁻³]). A composite potency value that combined LLNA and human potency data was defined, which aided the performance of the proposed model for the prediction of NESILs. The potency model accurately predicted sensitizing potency, with cross-validation errors of 2.75 and 3.22 fold changes compared with NESILs from LLNAs and humans, respectively. In conclusion, the results suggest that the GARDskin Dose–Response assay may be used to derive an accurate quantitative continuous potency estimate of skin sensitizers.

Recent years have seen a surge in development of alternative methods for assessment of skin sensitizers. This has been in order to meet the regulatory and industrial demand for accurate safety assessment without the use of animal experimentation, and to address societal expectations. Today, 3 such assays are formally validated and have achieved OECD Guideline status (EC, 2013, 2014, 2015). However, partly due to insufficient predictive capacity, none of the validated assays have received a recommendation for use as a stand-alone method. In addition, there is a view that accurate hazard identification must be based upon an assessment of several key events in the relevant Adverse Outcome Pathway (AOP).

Proactive identification and characterization of hazards attributable to chemicals are central aspects of risk assessments. Current legislations and trends in predictive toxicology advocate a transition from in vivo methods to nonanimal alternatives. For skin sensitization assessment, several OECD validated alternatives exist for hazard identification, but nonanimal methods capable of accurately characterizing the risks associated with sensitizing potency are still lacking. The GARD (Genomic Allergen Rapid Detection) platform utilizes exposure-induced gene expression profiles of a dendritic-like cell line in combination with machine learning to provide hazard classifications for different immunotoxicity endpoints. Recently, a novel genomic biomarker signature displaying promising potency-associated discrimination between weak and strong skin sensitizers was proposed. Here, we present the adaptation of the defined biomarker signature on a gene expression analysis platform suited for routine acquisition, confirm the validity of the proposed biomarkers, and define the GARDpotency assay for prediction of skin sensitizer potency. The performance of GARDpotency was validated in a blinded ring trial, in accordance with OECD guidance documents. The cumulative accuracy was estimated to 88.0% across 3 laboratories and 9 independent experiments. The within-laboratory reproducibility measures ranged between 62.5% and 88.9%, and the between-laboratory reproducibility was estimated to 61.1%. Currently, no direct or systematic cause for the observed inconsistencies between the laboratories has been identified. Further investigations into the sources of introduced variability will potentially allow for increased reproducibility. In conclusion, the in vitro GARDpotency assay constitutes a step forward for development of nonanimal alternatives for hazard characterization of skin sensitizers.

Many biological data acquisition platforms suffer from inadvertent inclusion of biologically irrelevant variance in analyzed data, collectively termed batch effects. Batch effects can lead to difficulties in downstream analysis by lowering the power to detect biologically interesting differences and can in certain instances lead to false discoveries. They are especially troublesome in predictive modelling where samples in training sets and test sets are often completely correlated with batches. In this article, we present BARA, a normalization method for adjusting batch effects in predictive modelling. BARA utilizes a few reference samples to adjust for batch effects in a compressed data space spanned by the training set. We evaluate BARA using a collection of publicly available datasets and three different prediction models, and compare its performance to already existing methods developed for similar purposes. The results show that data normalized with BARA generates high and consistent prediction performances. Further, they suggest that BARA produces reliable performances independent of the examined classifiers. We therefore conclude that BARA has great potential to facilitate the development of predictive assays where test sets and training sets are correlated with batch.