Curie Brief
Turn on cookies to sign in
Signing in saves your progress to your Curie account. We can only do that with cookies on — turn them on to continue.
News
Stanford researchers have built two groundbreaking AI models — including TranscriptFormer — trained on 112 million cells from 12 species to decode gene expression patterns. The models can compare cells across species, distinguish healthy from diseased cells, and may one day guide the design of entirely new cell-based therapies. Researchers call it the start of "a whole new field."
Stanford Medicine researchers have unveiled two AI models that could fundamentally change how scientists study cell biology. The first, called Universal Cell Embedding, laid the groundwork for a more powerful second-generation model, TranscriptFormer — trained on data from 112 million cells spanning 12 species, from yeast to humans. Think of it like ChatGPT, but instead of predicting missing words, it predicts missing gene expression values.
The result is a "universal mathematical space" where cells from any organism can be placed and compared. That means scientists can now explore evolutionary relationships, identify cell types in newly studied species, and distinguish healthy cells from diseased ones — all at a scale the human brain simply can't process alone.
By the Numbers:
Why it matters: These models don't just crunch big data — they open doors to discovering new disease mechanisms and designing novel cell-based treatments. As the researchers put it, this is the beginning of a decade-long new field in biology.
