Treffer: Development and evaluation of a novel consensus in silico model within the OECD GL497 defined approach for skin sensitization prediction.

Skin sensitization is a key endpoint in chemical safety assessment, involving multiple key events (KEs) in the adverse outcome pathway. The integrated testing strategy defined approach (ITS-DA) in OECD GL497 combines non-animal methods addressing KE1 and KE3 with an in silico component to predict sensitization. We developed a novel consensus in silico model integrating a rule-based system with metabolic simulation and two statistical models trained on local lymph node assay (LLNA) and guinea pig maximization test (GPMT) data. When applied as the in silico element of the ITS-DA, the model achieved balanced accuracies of 77 % for LLNA and 70 % for human data, surpassing existing in silico approaches. Furthermore, we developed a rule- and statistics-based KE1 replacement model (RSRKE1) that estimates protein-binding potential without in chemico KE1 assays. Combining RSRKE1 with the human Cell Line Activation Test (h-CLAT, KE3) and the consensus model yielded sensitivities of ∼80 % and potency classification accuracies of ∼70 % for LLNA, with similar performance for human data. These results demonstrate that the consensus model and RSRKE1 can maintain predictive performance comparable to established ITS-DA workflows, while eliminating the need for certain in vitro assays. This approach supports regulatory acceptance of non-animal testing strategies for skin sensitization assessment.
(Copyright © 2026 Elsevier Inc. All rights reserved.)

Declaration of competing interest I have nothing to declare.