A new generative AI for biology can fill in the blanks of missing protein structures and suggest ways to fix genetic mutations.
April 29, 2026
Original Paper
MIMIC: A Generative Multimodal Foundation Model for Biomolecules
arXiv · 2604.24506
The Takeaway
Most biological AI models focus on one thing, like predicting how a protein folds. This foundation model, called MIMIC, looks at the entire biological system, including DNA and proteins, to find hidden patterns. It can identify exactly where a mutation has gone wrong and propose the specific edits needed to restore health. This is like having a predictive text tool for the code of life that can fix typos in our genome. The ability to generate missing biological data across different scales could drastically speed up the development of personalized medicine. It moves AI from just observing biology to actively suggesting repairs.
From the abstract
Biological function emerges from coupled constraints across sequence, structure, regulation, evolution, and cellular context, yet most foundation models in biology are trained within one modality or for a fixed forward task. We present MIMIC, a generative multimodal foundation model trained on our newly curated and aligned dataset, LORE, linking nucleic acid, protein, evolutionary, structural, regulatory, and semantic/contextual modalities within partially observed biomolecular states. MIMIC use