Breakout Session Leaders: Malte Luecken & Maria Brbic

This session will examine the opportunities and challenges involved in developing an HCA foundation model capable of supporting a broad range of biological and translational applications. Participants will discuss the key tasks and benchmarks such models should be evaluated against, as well as the standards for robustness, interpretability, scalability, and long-term utility that the HCA community should expect. The conversation will also explore how to balance scientific rigor and real-world applicability, including questions around when data quality is “good enough” to support model development versus waiting for more complete or ideal datasets. In addition, the session will consider whether the HCA should focus on creating a shared base model alone or also support fine-tuned models optimized for specific downstream tasks and community needs. Finally, participants will discuss strategies to encourage large-scale data sharing and re-use, recognizing that successful foundation models will require unprecedented levels of interoperable, high-quality data contributed across the HCA ecosystem.