MIT’s IRIS model reads the signals steering developing cells

MIT researchers say a machine-learning model called IRIS can infer which chemical signals cells received during development, including in cell types it was not trained on. That could help scientists map how tissues form and guide stem cells towards specific roles, though it is a research tool, not a treatment.

Watch Desk analysis

What happened

MIT News describes IRIS as a neural network trained on gene-activity measurements from human embryonic stem cells exposed to combinations of six signalling pathways. The model looks for pathway-specific patterns, or “fingerprints”, in gene activity.

The researchers then tested it on individual cells from developing mouse embryos. MIT News says IRIS predicted where and when signals were active in cells destined to become heart, gut, muscle and spinal-cord tissue. In a further test, the model predicted a signal that could encourage lung-specific development, and experiments in mouse embryos confirmed that prediction. The detailed findings appeared in Nature Methods on 8 September, according to MIT News.

Why it matters

Researchers have often had to map how a pathway works separately in each cell type. If these fingerprints hold across cell types, IRIS could make it easier to reconstruct a cell’s signalling history without repeating that work for every one.

That may help researchers investigate development and disease, or make more accurate lab-grown tissue models for testing potential therapies. It does not show that the model can produce working organs or treatments; those remain possibilities, not results.

Our read

The interesting step is not simply that a model predicts gene activity. It is that the researchers report it can carry patterns learned in human stem cells into different cell types, then help point to a signal worth testing. The mouse experiments give that prediction a useful check, while leaving a long road between a promising lab model and a clinical application.

What to watch

• Whether further experiments reproduce the pathway fingerprints across more cell types and developmental stages.
• How well IRIS predicts signals beyond the datasets and pathways used in this study.
• Whether the approach improves the reliability of lab-grown tissue models in practice.

Discussion spark: If a model can predict which signals steer cells towards a particular tissue, what evidence should researchers need before using those predictions to guide stem-cell engineering?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.