Controlling Biology with Fabian Theis
AI Summary
Fabian Theis from Helmholtz Munich discusses the evolution of Virtual Cell concepts, multi-scale biological data, and the challenges of perturbation prediction. Despite historical setbacks where simple baselines outperformed complex models, improved evaluation metrics and massive new data initiatives like the $1.8B Biohub push are revitalizing systems biology. These advancements combine advanced imaging, lab-in-the-loop automation, and multi-modal AI to model biological systems from molecules to organs.
Key Takeaways
- Fabian Theis leads multi-scale, multi-modal biological data modeling efforts at Helmholtz Munich.
- Recent massive funding initiatives ($1.8B from Biohub, NIH, DOE, Google, and Meta) aim to scale high-precision biological data collection.
- Deep learning models for perturbation prediction previously struggled against simple baselines due to poorly calibrated metrics, but proper calibration shows neural networks outperforming them.
Why it matters
Understanding multi-scale biological data and mastering perturbation prediction are crucial for advancing precision medicine, drug discovery, and computational biology, shifting AI from basic molecular design to holistic cell and tissue control.
Conversation 0
Be the first to share your perspective.