ONTOX Insights #7

Artificial intelligence is rapidly transforming how kidney toxicity risks are identified and managed. By linking patient data with patterns of chemical-induced kidney injury, new AI-based approaches are improving our ability to predict nephrotoxicity before severe damage occurs. Recent advances integrate large clinical datasets with molecular and toxicological information to better identify individuals at higher risk. These tools also help uncover the underlying biological mechanisms of kidney injury, supporting more precise and preventive care. Overall, AI-driven strategies are paving the way toward predictive nephrology and safer use of medicines and chemicals in vulnerable patient populations.

Chemical safety assessment still relies heavily on animal testing, despite ethical concerns and its limited ability to predict human outcomes. New Approach Methodologies (NAMs), including in vitro and in silico tools, are emerging as promising alternatives, but many computational models still lack strong biological and mechanistic grounding. Recent advances focus on understanding adverse effects through mechanistic frameworks such as Adverse Outcome Pathways (AOPs. Inspired by the Disease Maps Project, this paper introduces Physiological Maps (PMs) as comprehensive graphical representation of biochemical processes related to specific organ functions. PMs uses Systems Biology Graphical Notation (SBGN) and curation guidelines, supporting data analysis, education, and the development of more human-relevant in vitro and in silico models. By bridging toxicology and systems biology, PMs help advance next-generation, mechanism-based chemical risk assessment.

Adverse outcome pathways (AOPs) provide a structured framework for organizing scientific knowledge in regulatory toxicology, and describe how a chemical interaction at the molecular level can trigger a sequence of measurable biological events that ultimately lead to an adverse health outcome. AOPs consist of multiple key events that are linked through clearly defined causal relationships, supporting transparency and consistency in hazard assessment. By focusing on mechanistic understanding rather than apical animal endpoints, AOPs facilitate the use of human-relevant alternative methods.

In silico methods are becoming a key component of new approach methodologies (NAMs), offering efficient and human-relevant alternatives to animal testing in chemical safety assessment. This study introduces two Liver Physiological Maps (PMs), which are machine-readable, curated representations of the molecular mechanisms governing two major liver functions, specifically (i) The LiverLipidPM focuses on liver lipid metabolism, including detailed pathways of fatty acid synthesis, cholesterol metabolism, and lipid catabolism in hepatocytes, while (ii) the LiverBilePM focuses on bile acid biosynthesis, transport, and secretion processes between hepatocytes and cholangiocytes. Both maps integrate metabolic signaling pathways and regulatory networks and are available as interactive online tools that allow data visualization and linkage to biological ontologies. Comparative analysis shown unique mechanisms to each map and overlap with existing Adverse Outcome Pathway (AOP) networks. These developed liver PMs provide a valuable foundation for refining AOPs, identifying new molecular targets, and supporting the development of human-relevant in vitro and computational models for hepatotoxicity assessment.

This study reviewed clinical case reports of drug-induced fatty liver disease (DIFLD) to classify drugs according to distinct clinical phenotypes. Based on clinical, biochemical, and histological features, seven distinct clusters of drugs were identified, ranging from compounds with no steatotic effects to those causing severe steatohepatitis with cholestasis. Mild clusters primarily worsen existing metabolic steatosis, while more severe clusters are linked to mitochondrial dysfunction, inflammation, and impaired lipid export. By connecting clinical phenotypes with underlying toxicological mechanisms, this classification improves the diagnosis and risk assessment of DIFLD.

Follow us on social media!