• How do we systematically account for missing biological complexity, such as tissue interactions, pharmacokinetics and microenvironment effects, when extrapolating from in vitro trained AI models to in vivo outcomes?
• What role do multimodal data integration strategies, including imaging, omics and functional assay data, play in improving the robustness of in vitro to in vivo translation in AI models?
• How can we validate and benchmark AI model predictions to build confidence in their in vivo relevance for drug discovery?
• Which AIML approaches are most effective at modelling complex in vivo responses using predominantly in vitro training data?

(Discussion points are subject to change)