Regulatory authorities frequently need information on a chemical's capacity to produce acute systemic toxicity in humans. Due to concerns about animal welfare, human relevance, and reproducibility, numerous international initiatives have centered on finding a substitute for using animals in acute systemic lethality testing. These substitutes include the more current in-silico and in vitro techniques. Meanwhile, Advances in artificial intelligence and computational resources have led to a rise in the speed and accuracy of machine learning algorithms. Therefore, new approach methodologies (NAMs) based on in-silico modeling are considered a suitable place to start, even though many non-animal testing approaches exist for evaluating the safety of chemicals. Eventually, in this investigation, we have developed a hybrid computational model for acute inhalational toxicity data. In this case study, two major in silico techniques, QSAR (quantitative structure-activity relationship) and qRA (quantitative read-across) predictions, were utilized in a hybrid manner to extract more insightful information about the compounds based on similarity as well as the physicochemical properties. The findings of this investigation demonstrate that the integrated method surpasses the traditional QSAR model in terms of statistical quality for inhalational toxicity data, with greater predictability and transferability, due to a much smaller number of descriptors used in the hybrid modeling process. This hybrid modeling technique is a promising alternative, which can be paired with other methods in an integrated manner for a more rational categorization and evaluation of inhaled chemicals as a substitute for animal testing for regulatory purposes in the future.
Keywords: Acute systemic toxicity; Hybrid model; Inhalational toxicity; NAMs; QSAR; qRA; qRASAR.
Copyright © 2024 Elsevier Ltd. All rights reserved.