MIT researchers develop IRIS AI model to decode cell communication language
Researchers led by MIT assistant professor of biology Pulin Li and graduate student Nicholas Hutchins have developed a machine learning model called IRIS to decode the signaling pathways cells use to communicate. Published in Nature Methods on Sept. 8, the neural network-based model identifies unique gene activity fingerprints that remain consistent across different cell types, enabling scientists to reconstruct signaling histories and potentially accelerate stem cell engineering, organoid creation, and drug testing.
Key points
- MIT researchers created a machine learning model named IRIS to decode cell communication signaling pathways.
- The model identifies consistent gene activity fingerprints across different cell types.
- Findings were published in Nature Methods on Sept. 8.
- IRIS successfully predicted signaling pathway activation in mouse embryo cells during gastrulation.
What happened
Researchers at MIT and the Whitehead Institute for Biomedical Research have developed a neural network-based artificial intelligence model called IRIS designed to decode how cells communicate with one another during development. Led by MIT assistant professor of biology Pulin Li and graduate student Nicholas Hutchins, the team discovered that each cellular signaling pathway leaves behind a unique fingerprint of gene activity that remains consistent across different cell types.
The open-access findings were published in the journal Nature Methods on Sept. 8.
What changed
Traditionally, scientists assumed that the effects of signaling pathways varied widely across cell types, requiring arduous separate mapping for each individual cell type. The IRIS model overcomes this limitation by using transfer learning to detect pathway fingerprints and reconstruct signaling histories across multiple cell types, even those it has not previously encountered.
The research team trained IRIS on an experimental dataset measuring how human embryonic stem cells responded to combinations of six major signaling pathways. When tested on single cells from mouse embryos during gastrulation, IRIS accurately predicted signaling pathway activation in cells forming heart, gut, muscle, and spinal cord tissue, and successfully predicted signals needed to encourage lung-specific development.
Availability
The detailed findings are available open-access through the journal Nature Methods following their publication on Sept. 8.
What is next
Researchers suggest the IRIS model could accelerate stem cell engineering for regenerative medicine and improve the creation of 3D organoid models for studying diseases such as asthma, lung cancer, and pulmonary fibrosis. Scientists plan to use these insights to better control stem cell fate and test potential therapies in lab settings.
Why it matters
The IRIS AI model allows researchers to reconstruct cell signaling histories across different cell types without mapping each type separately, potentially accelerating stem cell engineering and disease research.
What we know
- Researchers at MIT and the Whitehead Institute developed an AI model called IRIS to identify gene activity fingerprints.
- The findings were published in Nature Methods on Sept. 8.
- IRIS was trained on human embryonic stem cell data and tested on mouse embryo cells during gastrulation.
What remains unclear
- Whether IRIS will successfully accelerate stem cell engineering and organoid creation across broader clinical and industrial applications remains a forecast.
