Despite rapid advances in co-folding models, foundation architectures, and AI-driven structure prediction, one constraint continues to limit progress: access to sufficiently diverse, high-quality structural datasets. Individual organizations rarely hold enough breadth to train highly predictive models across modalities and target classes, yet historical consortium efforts have struggled due to confidentiality concerns, governance complexity, and unclear value exchange.

This interactive discussion will allow open discussion with industry peers to identify opportunities for

collaboration by discussing:

Data Contribution:

• The minimum viable data contributions (structural diversity, ligand space, experimental annotation) that would be required to improve predictive models
• How value should be quantified and fairly exchanged?
• Whether collaborations should be modality-specific?

Competitive Boundaries:

• What technical, legal and operational safeguards would be required?
• What would need to change from previous consortium models to make future participation viable and more attractive?