How the IRIS AI Model Decodes Cell Signaling Pathways

How does the IRIS AI model analyze cell signaling pathways?

IRIS is a new computational approach developed at MIT that decodes the language cells use to communicate. It functions as a neural-network-based model that examines a cell’s overall gene activity to estimate which signaling pathways were active during development.

The model identifies fingerprints of various signaling pathways to determine the specific signals a cell received at different stages, even in cell types it has not encountered previously. IRIS is a specialized tool for biology.

What role does gene activity analysis play in IRIS?

As an embryo develops from a small cluster of stem cells, these cells transition from a blank slate into specialized types like brain, liver, or muscle cells. This process organizes them into three-dimensional structures including tissues and organs.

The fate of a cell depends on which genes are turned on or off, as these patterns of gene activity dictate the cell’s structure and function. To coordinate this, cells constantly send and receive chemical signals to and from their neighbors.

These signals help cells understand their location, their developmental stage, and their ultimate role. The messages travel through multi-step sequences known as signaling pathways, which translate external information into specific changes in gene activity.

Why was decoding cell signals difficult before?

Retracing the sequence of instructions a cell receives is a powerful way to understand tissue development and disease, but scientists have long faced significant challenges. Researchers previously assumed that the effects of signaling pathways varied widely across different cell types, which would require mapping each pathway separately for every type of cell.

While this work progresses, individual researchers like senior Veronica Calzadilla Sosa have followed their own long paths toward their MIT degrees with the support of their community.

How MIT AI Research Advances Developmental Biology

The IRIS model functions by identifying a unique “fingerprint” for each signaling pathway, which is a specific pattern of gene activity reflecting the signals a cell has received. Because these fingerprints remain consistent across different cell types for the same pathway, the model can reconstruct the signaling history of a cell regardless of its specific type. This approach marks a major shift from traditional research, which requires scientists to experimentally test every single pathway in every possible cell type.

To achieve this, researchers Pulin Li and Nicholas Hutchins employed a technique called transfer learning. Li explains that this is similar to voice recognition systems like Siri, which are trained primarily on English but apply that knowledge to recognize other languages. By applying this principle, IRIS can effectively operate across many different cell types.

The model was trained on a large dataset measuring how thousands of human embryonic stem cells responded to dozens of combinations of six major signaling pathways at various developmental stages. This training created a comprehensive map, or atlas, that shows how these signaling combinations influence cell behavior. This allows for the mapping of signaling histories of every cell within a mouse or human embryo at an unprecedented scale.

For instance, researchers like Benjamin Lindquist study how injecting randomness into early AI can lead to models that sometimes hallucinate. Other developments at the institute include “X-Hinges,” which helps design 3D-printed items with sensors that adapt to human contact.

How was the model validated?

The research team validated the IRIS model using single cells from mouse embryos during gastrulation. The model successfully predicted when and where specific signaling pathways would activate in cells destined to become heart, gut, muscle, and spinal cord tissues.

When the team applied IRIS to identify signals required for a cell type critical to lung development, the model predicted that activating a specific pathway would drive that development. Experiments in mouse embryos later confirmed these predictions. Regarding the broader research environment, MIT President Sally Kornbluth recently stated that the university has never needed guidance and insight more than it does now.

What are the practical applications for medicine?

This computational approach can accelerate stem cell engineering, improve disease models, and inform medical therapies. By improving the creation of organoids—miniature 3D models that mimic real organs—the technology helps researchers study disease mechanisms and test new drugs.

These advanced models enable the study of conditions such as asthma, lung cancer, and pulmonary fibrosis. Researchers can use this data to test potential therapies and design regenerative treatments aimed at repairing damaged tissue. While these innovations help meet civic and cultural needs, a legal scholar noted in the Compton Lecture that there is a potential risk of a "loss of the Enlightenment" that must be balanced against such technological progress.

Is the research peer-reviewed?

The release of this study is currently a preprint or press release. It has not yet undergone the formal peer-review process.