Session 4: Develop Skin Sensitization ITS

Combining Data Streams and Modeling Approaches to Develop Skin Sensitization ITS
Nicole Kleinstreuer, ILS/NICEATM

Nicole Kleinstreuer, Ph.D.
Director, Computational Toxicology Group




The National Toxicology Program Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM) is using the OECD adverse outcome pathway (AOP) for skin sensitization to develop integrated testing strategies (ITS) that will accurately identify potential skin sensitizers.

One project is a Bayesian network model built by NICEATM in collaboration with Procter and Gamble using R, an open-source statistical computing language, that provides probabilistic predictions of skin sensitization potency based on in silico and in vitro information as well as skin penetration/bioavailability characteristics.

Another project with the Interagency Coordinating Committee on Validation of Alternative Methods (ICCVAM), is building an ITS using validated in vitro (h-CLAT and Keratinosens), in chemico (DPRA), and in silico (OECD Toolbox) test methods, along with physicochemical parameters that could influence skin penetration, to predict skin sensitization hazard.

These data were used as input features in simple battery approaches and more complex machine learning models to predict the murine local lymph node assay (LLNA) results, and the best models achieved over 96% predictive accuracy on both the training and the test sets.

Additional work examining the ToxCast and Tox21 high-throughput screening (HTS) data sets identified a number of HTS assay targets that mapped to the AOP for skin sensitization and were correlated with LLNA results.

Additional ITS are being developed to predict human sensitization potential, including potency classification, and that will incorporate HTS assays and structural features together with validated in vitro assays and physicochemical properties.