Skip to main content
Contact Info
Ghada Tagorti
OpenTox Intern

OpenTox Virtual Conference 2023

Prediction of Molecular Initiating Event (MIE) Inceptive Stage in Adverse Outcome Pathway (AOP) Using Large Language Model

The Adverse Outcome Pathway (AOP) is a conceptual framework encompassing molecular initiating events (MIE) followed by key events (KEs), enabling the utilization of alternative assays in toxicology regulations. The inceptive stage of this framework involves an MIE which activates cellular signaling pathways as a result of direct interaction of a compound with a biological target resulting in toxic clinical manifestations. In most cases, these compounds exhibit promiscuous binding and interaction profiles with various molecular targets, thus identifying the molecular initiating events responsible for toxicity induction is a challenging task. In this study we applied AI-driven literature search tools; EPA’s Adverse Outcome PathwayDatabase (AOP-DB) and AOP-helpFinder 2.0 web server to characterize chemical toxicity of compounds through the concept of AOPs and MIEs. A total of 25 compounds with 827 AOPs were identified. Thereafter data was selected for pre-trained modeling and tokenization. This was then fine-tuned by defining training arguments such as output directory, batch size, training steps, and learning rate. AOP framework, which places emphasis on Molecular Initiating Events and their downstream Key Events has revolutionized our approach to toxicology, and with the use of a large language model, our prediction of compound toxicity can further be enhanced. This study depicts how machine learning can enhance our understanding of biological processes and improve our capacity to elucidate the challenges behind the complexity of toxicological compounds.