Understanding RNA Foundation Models: A Primer

Foundation models have transformed natural language processing and computer vision. Now, they're beginning to do the same for biology. But what makes an RNA foundation model different from a large language model — and why does it matter?

What Is a Foundation Model?

A foundation model is a large-scale AI system trained on broad data that can be adapted to a wide range of downstream tasks. GPT-4 is a foundation model for text. ORIGIN™ is a foundation model for RNA.

Why RNA Needs Its Own Foundation Model

RNA data has unique statistical properties that generic tabular models fail to capture. Cell-free RNA is extremely sparse — most genes show zero expression in any given sample. The few that do show expression follow heavy-tailed distributions. And genes don't act independently; they form complex correlation networks driven by shared tissue sources and biological pathways.

Training a foundation model on synthetic data that doesn't reflect these properties produces models that fail on real cfRNA. That's why Eigen Bio developed cfRNA-specific approaches grounded in biological causal models.

What Can RNA Foundation Models Do?

• Cancer detection: Identify early-stage cancers from a blood draw
• Drug target discovery: Find therapeutic targets directly from cfRNA patterns
• Treatment monitoring: Track molecular response to therapy non-invasively
• Biological simulation: Predict how systems respond to perturbations

The era of RNA foundation models is just beginning — and the potential impact on human health is enormous.