Ladeira, L.L.LadeiraMazein, A.A.MazeinOstaszewski, M.M.OstaszewskiVerhoeven, A.A.VerhoevenKuchovská, E.E.KuchovskáSanz-Serrano, J.J.Sanz-SerranoDrees, A.A.DreesReiche, KristinKristinReicheSewald, K.K.SewaldFritsche, E.E.FritscheStaumont, B.B.StaumontGeris, L.L.GerisVinken, M.M.Vinken2024-11-212024-11-212024-09https://publica.fraunhofer.de/handle/publica/47915510.1016/j.toxlet.2024.07.164Adverse Outcome Pathways (AOPs) serve as frameworks connecting molecular initiating events to adverse outcomes through key events (KE), which are essential for understanding the link between chemical exposure and adverse health effects. Current graphical representations of AOPs are being adapted for machine-readability, aligning with FAIR principles (Findability, Accessibility, Interoperability, and Reuse of digital assets). The Systems Biology Graphical Notation (SBGN) elements enhance AOPs with standardized graphical notation, improving visualization and interpretability, as introduced by Mazein and collaborators (2023). In addition, artificial intelligence (AI)-based systematic review, data screening and curation have been performed by van Ertvelde and collaborators (2023) to accelerate the building of large AOP networks. The KE descriptor concept overlaps with the biology enrichment present in Mazein's work, linking biological entities to biological activity (as KE) and ultimately enhancing their mechanistic representation and relevance. This work aims to strengthen the bridge between toxicology and systems biology by proposing a semi-automated workflow for AOP network development. In this sense, we developed a pilot study on liver steatosis, addressing challenges and opportunities in developing SBGN-based AOP networks and proposing scalable solutions that can be adjusted to meet various demands. This approach integrates different data acquisition methods with automated network construction, followed by manual graphical improvements. Our workflow consists of: a) data checking, annotation and disambiguation; b) automated data processing; c) conversion into a CellDesigner SBML network (Funahashi et al., 2008), using R scripts applying functions from the minervar package (Gawron et al., 2023); d) manual design editing for improved layout and human interpretability. The MINERVA platform (Hoksza et al., 2020) was used for automated annotation, sharing, visualization, and exploration of the network. This approach was applied to extended case-studies with different data acquisition methods (i.e., AI-driven data extraction, literature review and integration of AOPs from AOP-Wiki) highlighting its versatility. The proposed approach leverages established standards and automated methods to expedite machine-readability and ensure FAIR principles compliance in AOP networks. Utilizing the proposed workflow in constructing AOP networks not only boosts reproducibility and interoperability but also facilitates the development of more accurate and biologically relevant networks. The incorporation of KE descriptors and biology enrichment simultaneously expands mechanistic relevance, improving the overall accuracy and comprehensiveness of the AOP networks.enjournal article