Causal biological network models consisting of multiple biological pathways involved in a given biological process can serve to contextualize gene expression changes and unravel key mechanisms responsible for those changes. The transcriptomic data from the respiratory nasal epithelium (RNE) of rats exposed to formaldehyde have been investigated using such causal biological network models. The resulting association between the biological impact assessed by network perturbation and the squamous cell carcinoma rate in the RNE after two years has been further investigated to gain mechanistic insights. A detailed node-level investigation revealed that while similar network models were impacted across exposure doses, the directionality of the effect was opposite for the lowest doses compared to high doses. In particular, NF-κB was inferred to be upregulated in response to the two higher doses and downregulated in response to the lower doses in the context of the epithelial innate immune activation network model. This highlighted a dose threshold indicative of a long-term biphasic effect of formaldehyde exposure leading to carcinogenicity. The presented approach could be used to establish the mechanism of action or grouping of compounds based on impacted regions in the network models.