{
  "id": 14788549,
  "title": "AI model decodes the language cells use to communicate",
  "url": "https://urgent.news/2026/10/15/ai-model-decodes-the-language-cells-use-to-communicate-14788549",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-15T20:50:00.000Z",
  "source": {
    "name": "MIT News AI",
    "slug": "mit-news-ai",
    "url": "https://news.mit.edu/2026/ai-model-decodes-language-cells-use-communicate-1015"
  },
  "original_language": "en",
  "account": "Embryonic cells initially serve as a blank slate, gradually adopting specialized roles like those of brain, liver, or muscle cells, before assembling into three-dimensional tissues and organs. Each cell's fate hinges on the genes activated or suppressed, dictating its structure and function. This cell-to-cell communication is facilitated by chemical signals exchanged with adjacent cells, allowing them to comprehend their location, developmental stage, and intended role. These signals travel through multi-step pathways, converting external cues into specific gene activity changes within the cell. For researchers, tracing the sequence of instructions a cell has received promises valuable insights into tissue development and disease mechanisms. However, this endeavor has been hindered by the assumption that signaling pathway effects vary significantly among cell types, necessitating separate pathway mapping for each type—an onerous task. MIT assistant professor Pulin Li and graduate student Nicholas Hutchins have now revealed that each pathway leaves a unique fingerprint—a distinct pattern of gene activity specific to the signals it has encountered. This fingerprint remains consistent across cell types for the same pathway, enabling scientists to reconstruct signaling histories across numerous cell types without the need for separate mappings. This breakthrough was achieved using a machine learning model called IRIS, which can discern signaling pathway fingerprints and pinpoint the signals cells received at various developmental stages, even for previously unencountered cell types. This AI-driven approach marks a significant leap over traditional methods that necessitate experimental testing of every pathway in every cell type. Described in Nature Methods on September 8, these findings could accelerate stem cell engineering in regenerative medicine and enhance the creation of organoids—miniature 3D organ models—to study disease mechanisms and drug testing. By understanding the signaling patterns that dictate a stem cell's fate, researchers can replicate these signals to control stem cells in lab or medical settings. The IRIS model, trained on a dataset measuring how human embryonic stem cells responded to six major signaling pathways at different developmental stages, provides a comprehensive \"atlas\" of how signaling combinations influence cell behavior. Tested on mouse embryos during gastrulation, the model accurately predicted signaling pathway activation in cells destined to become heart, gut, muscle, and spinal cord tissue. With IRIS, scientists can now not only comprehend cell-to-cell communication fundamentals but also devise a practical strategy for guiding stem cells into specific functional types in the lab. For instance, when IRIS identified the signals required for lung-specific development, experiments in mouse embryos validated the model's prediction. By determining the precise signal combinations that drive lung cell development, researchers can reliably generate precise lab-grown lung tissue models. This enhances the study of lung-related diseases like asthma, cancer, and fibrosis, potentially leading to therapies that repair damaged lung tissue.",
  "summary": "A new computational approach reveals how cells communicate to assemble into tissues. This could accelerate stem cell engineering, improve disease models, and inform therapies.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "MIT News Research",
        "title": "AI model decodes the language cells use to communicate",
        "url": "https://urgent.news/2026/10/15/ai-model-decodes-the-language-cells-use-to-communicate",
        "published": "2026-10-15T20:50:00.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}