Report on pre-validation of an animal-free alternative method (NAM) for regulatory safety testing: InfiniteLungDT, an in-vitro-learned digital twin for the prediction of material-triggered chronic neutrophilic lung inflammation

Abstract

Until now, there has been no animal-free alternative method for predicting chronic inflammation and delivering the associated dose responses, the timing of onset, and the duration of inflammation, as required by regulatory agencies. We present the results of pre-validation of an in-vitro-learned digital twin (InFiniteLungDT) capable of predicting chronic neutrophilic lung inflammation for regulatory use. The method is based on measuring the dynamics of early biological effects in vitro induced by respirable materials or their mixtures, without the need to know their intrinsic properties. We constructed the digital twin(s) for each of the material, for which we have in vivo exposure data. The instillation data set, comprising 49 different nanomaterials, was used as the primary anchor to calibrate the model. Inhalation data set, comprising 7 different nanomaterials, compliant with OECD TG 412, was used to show the general applicability of the method across species and for different exposure scenaria. In total, about 3094 single mouse exposures and 364 rat exposures (and approx. 775/225 non-exposed mouse/rat controls) were used to predict concentration-dependent time-evolved neutrophil influx into the lung. The accuracy (predictive capacity) of LOAEL determination is 93% for instillation and 84% for inhalation exposure. Taking into account the time-to-deliver-result being less than 1 week, this proves that the effect of inhaled material from acute to chronic conditions can be assessed orders of magnitude faster and cheaper than in a reference animal study.