Fabian Theis on Virtual Cells and AI Biology
Understand the story
On November 6, 2026, the Latent Space podcast released a conversation with Fabian Theis of Helmholtz Munich on controlling biology and the next hard problems in Virtual Cell modeling. In the interview, Theis discussed the current state of Virtual Cell modeling: in the 2025 Arc Virtual Cell Challenge, many complex models performed worse than even a constant baseline on the Perturbation Prediction task; but with better-calibrated evaluation metrics, deep learning models can come out ahead. There is currently only this one report on the event, with no follow-up coverage yet.
AI overview of reporting · Updated 55 min前
Reporting timeline
Explore different aspects of the story through its reporting.
- Latent SpaceControlling Biology - Fabian Theis, Helmholtz Munich
Fabian Theis of Helmholtz Munich discusses the state of Virtual Cell modeling on the Latent Space podcast: in the 2025 Arc Virtual Cell Challenge, many sophisticated models failed to beat even a constant baseline on Perturbation Prediction, but with better-calibrated evaluation metrics, deep learning models can win.
Story heat trend
Not enough continuous observations to show a trend yet.