TOXTRUST: a tool leveraging the Dempster-Shafer Theory for robust integration of NAM results in decision-making considering uncertainty

In the search for alternatives to replace in vivo studies, the application of assessment frameworks involving New Approach Methodologies (NAMs) often leads to the simultaneous availability of multiple pieces of evidence of different quality. When integrated into defined data structures, combinations of NAMs can generate answers to complex toxicological questions. However, if they are not integrated correctly, a collection of NAM results may produce misleading results that complicate the assessment process or lead to wrong conclusions. To support transparent decision-making in situations in which multiple NAMs are applied to generate results for the same toxicological question, we developed TOXTRUST ( www.github.com/phi-grib/TOXTRUST ) — an open-source computational tool integrating the mathematical framework of the Dempster-Shafer Theory (DST). In this article, we briefly describe the DST framework for the integration of NAM results to define the scope of its application and data requirements. This is followed by a description of the functionalities and infrastructure of TOXTRUST. Lastly, we illustrate how TOXTRUST can be applied to any endpoint with binary end-results, with a focus on the generation and interpretation of results expressed through probability bounds.