Researchers Developed Neural Network to Map Gene Pathways

The Whitehead Institute team created a tool called IRIS to predict signaling activity in embryonic cells.

Updated on Sept. 26, 2026 in Biotech

Isometric editorial illustration of geometric cellular structures and protein signaling pathways within a petri dish. — Researchers at the Whitehead Institute have developed IRIS, a neural network that predicts gene signaling pathways in embryonic cells to reduce reliance on animal testing. AI Illustration. Upload story photo >

Live Poll

Do you trust that AI simulations can accurately replace traditional animal research methods?

Researchers at the Whitehead Institute have developed a neural network named IRIS that estimates active gene signaling pathways within cells. The model allows scientists to analyze gene activity profiles to identify cellular responses without relying on traditional animal experiments.

Why it matters

This technology accelerates the identification of signaling histories in developing cells, offering a faster alternative to conventional animal testing. By predicting pathway activity through gene analysis, researchers can better understand how different cell types respond to developmental cues.

The IRIS neural network analyzes a cell's entire gene activity profile to determine active pathways. The team successfully utilized this model to generate 96 distinct cell states from a single experimental screen.

The players

Pulin Li

The lead researcher at the Whitehead Institute who spearheaded the development of the IRIS neural network.

Whitehead Institute

A world-renowned biomedical research institution affiliated with MIT where the study was conducted.

The details

The team trained IRIS on stem cells exposed to combinations of six signaling pathways, enabling it to map how cells respond to specific developmental signals. This capability successfully predicted signaling activity in nerve tissue and human airway cells, and helped the researchers establish a more effective protocol for growing lung tissue.

Screens on human embryonic stem cells were performed over 4-8 days.
Mouse embryo cells in the public atlas were aged 6.5-8.5 days post-fertilization.
Mouse embryos used in the lung tissue study were 9 days old.
Foregut tissue was maintained in a dish for a duration of 24 hours.
The standard lung cell protocol was completed in 7 days.

The Big Picture

This study follows a pattern set by the Whitehead Institute's developmental biology research program by integrating advanced computational modeling into traditional cellular study. The breakthrough shifts the scientific paradigm by moving away from reliance on animal models for mapping complex cellular signals.

For researchers and biotech developers, this tool offers a more efficient workflow for tissue engineering and regenerative medicine studies. The ability to model cell responses faster may ultimately reduce the time required to develop new therapeutic cell protocols.

The takeaway

The development of IRIS demonstrates that neural networks can significantly compress the timeline of biological research by predicting cell signaling outcomes. Researchers can implement this computational approach to reduce experimental overhead and accelerate the engineering of specific human tissue types.