What Is a Biological Foundation Model?
A biological foundation model is built on the idea that biology contains patterns that can be learned at scale. Instead of training a model for one narrow task, researchers train it on broad biological data, such as protein sequences, structures, gene expression, single-cell data, images, or multimodal lab measurements. The model can then be adapted for tasks like variant interpretation, protein design, target discovery, or disease modeling.
Plain-English explanation
A biological foundation model is built on the idea that biology contains patterns that can be learned at scale. Instead of training a model for one narrow task, researchers train it on broad biological data, such as protein sequences, structures, gene expression, single-cell data, images, or multimodal lab measurements. The model can then be adapted for tasks like variant interpretation, protein design, target discovery, or disease modeling.
Why it matters
Rare diseases often suffer from small datasets. A foundation model may help by transferring patterns learned from broader biology into rare contexts. That does not replace disease-specific validation, but it may help researchers generate better hypotheses and connect a rare disorder to pathways or mechanisms studied elsewhere.
How it works
The model learns statistical relationships in biological data, such as how protein sequences relate to structure or how gene-expression patterns relate to cell state. Researchers can then prompt, fine-tune, or evaluate the model for a specific question. Any useful prediction still needs experimental confirmation, especially when the disease is rare or poorly represented in training data.