72 Biological relevance of In finite™ Lung digital twin for animal/hypothesis-free prediction of material-induced chronic lung inflammation

Abstract Traditional animal toxicity test methods can be of limited biological relevance to human health effects1. Moreover, they are prohibitively costly, ethically problematic, and time-consuming. New Approach Methodologies (NAMs) aim to replace them by measuring mechanistic and cellular effects early in the pathway, thereby predicting human-relevant downstream effects. However, currently available NAMs are unreliable in modelling repeated-dose toxicity, including chronic toxicity, and exposure to different scenaria, which prevents broader application of NAMs in current regulatory risk assessment2. Reliable NAMs must generalize across all possible early modes of action (eMoAs)3. Relying on molecular-level data introduces excessive degrees of freedom. In contrast, cell-level functional responses reduce the problem to manageable dimensions and are in line with the concept of intermediate Key Events and IATA/AOPs as the basis for risk assessment, with the ultimate goal of replacing in vivo-centred AO observations. The In finite™ technology (InFiniteLungDT) overcomes all these challenges through high-throughput toxicology assessment of solid particles using (1) live-compatible interference-free imaging of epithelial and immune cell co-cultures, (2) pre-defined feature selection based on AOPWiki KEs and quantification algorithms, (3) automated translation of observables into eMoA networks via patented analysis5, and (4) long-term prediction evaluated against mice neutrophil influx datasets from about different 50 nanomaterials InfiniteLungDT predicts neutrophil influx with 82% accuracy, and identifies diesel exhaust particles caused phenotypic changes associated with increased cancer risk6. The system predicts long-term lung inflammation within one week, replacing year-long animal tests with a 50-fold faster, animal-free assessment.