Probabilistic hazard assessment for skin sensitization potency by dose–response modeling using feature elimination instead of quantitative structure–activity relationships

Abstract

Supervised learning methods promise to improve integrated testing strategies (ITS), but must be adjusted to handle high dimensionality and dose–response data. ITS approaches are currently fueled by the increasing mechanistic understanding of adverse outcome pathways (AOP) and the development of tests reflecting these mechanisms. Simple approaches to combine skin sensitization data sets, such as weight of evidence, fail due to problems in information redundancy and high dimension-ality. The problem is further amplified when potency information (dose/response) of hazards would be estimated. Skin sensitization currently serves as the foster child for AOP and ITS development, as legislative pressures combined with a very good mechanistic understanding of contact dermatitis have led to test development and relatively large high-quality data sets. We curated such a data set and combined a recursive variable selection algorithm to evaluate the information available through in silico, in chemico and in vitro assays. Chemical similarity alone could not cluster chemicals’ potency, and in vitro models consistently ranked high in recursive feature elimination. This allows reducing the number of tests included in an ITS. Next, we analyzed with a hidden Markov model that takes advantage of an intrinsic inter-relationship among the local lymph node assay classes, i.e. the monotonous connection between local lymph node assay and dose. The dose-informed random forest/hidden Markov model was superior to the dose-naive random forest model on all data sets. Although balanced accuracy improvement may seem small, this obscures the actual improvement in misclassifications as the dose-informed hidden Markov model strongly reduced "false-negatives" (i.e. extreme sensitizers as non-sensitizer) on all data sets.

Integrated Testing Strategies for Skin Sensitization Hazard and Potency Assessment—State of the Art and Challenges

“…ITS potency assessment approaches developed to date include 3-way and 4-way LLNA EC3 deterministic classification [19][20][21], pEC3 (molar equivalent of EC3) prediction [22][23][24], 4-way probabilistic EC3 classification [25][26][27] and 4-way probabilistic pEC3 classification with a possibility to estimate any percentile of pEC3 distribution [28]. All other ITS approaches mentioned in this mini-review are for hazard estimation only.…”

“…Approaches based on machine learning algorithms are very popular. Among them are linear regression regression-based methods [23,24,30] and nonlinear methods like neural networks [19,20], support vector machines [31] and random-forest models [27].…”

“…In a probabilistic framework transparent and consistent criteria to accept a prediction are easy to set up and the need for uncertainty factors is diminished. Despite calls for the type of approach [18] only two probabilistic ITS frameworks (Jaworska et al [25,26,28] and Luechtefeld et al [27]) have been published to date.…”

“…Among ITSs for skin sensitization performance evaluations based on an external dataset were published only for three approaches: Jaworska et al [24,25,28], Luechtefeld et al [27] and Strickland et al [31]. In addition to good statistical performance which is necessary, the ability to explain an input data conflict is an additional critical criterion to the successful use of ITS.…”

An economic approach to non-animal toxicity testing for skin sensitisation

“…We assume that testing has a 'value', if, and only if, the information revealed from testing triggers a welfare-improving decision on the use (or non-use) of a substance, compared to decision-making in the absence of additional information from testing. Thus, in contrast with information-theoretic approaches such as Bayesian Networks, Hidden Markov or quantitative Weight-of-Evidence approaches (Rorije et al, 2013;van der Veen et al, 2014a;Luechtefeld et al, 2015;Rovida et al, 2015), VOI analysis explicitly considers expected social gains and costs (called "payoffs") from any possible decision on the use of a substance, while accounting for the uncertainty of test information. Quantifying the VOI provides a tool which guides the choice and sequencing of methods in a testing strategy.…”