Jerry Achar

Although animals (e.g., rodents) are commonly used when testing potential harm caused by chemicals or drugs in humans, the use of animals is associated with high testing costs, is time-consuming, and has been challenged on ethical grounds. Promising non-animal testing alternatives, like in silico methods that aim to reduce, refine, or replace animal testing, have been proposed to potentially overcome these challenges. However, the acceptance and uptake of the methods in regulatory contexts are currently low, mainly due to uncertainties associated with their predictions. A more fundamental problem, however, is that the uncertainties are often not well understood nor transparently communicated to support the fit-for-purpose evaluation of in silico methods and to improve their utility in the regulatory predictors of potential harms caused by chemicals in human health. My research thus aims to systematically characterize uncertainties associated with in silico predictions in the chemical risk assessment of chemicals. I do this by identifying the sources of uncertainty within in silico modelling processes and analyzing whether the uncertainties are reducible or irreducible. I then develop a framework that systematically organizes the characterized uncertainties according to their sources and nature (reducible/irreducible). This framework's value will lie in its ability to improve scientific understanding of the manifestations of uncertainties related to in silico predictions and drive considerations to reduce them. I also contend that the framework could guide future studies to bring structure into characterizing different types of uncertainty related to emerging in silico methods and improve transparency for regulatory applications of the current and emerging in silico methods.