Scientists at Houston Methodist have developed an artificial intelligence platform that can decode how cells communicate inside the human body, opening a window into conditions such as Alzheimer’s disease, cancer and potentially accelerating the search for new treatments.
Cells in the human body are constantly “talking” to each other—sending signals that tell neighboring cells when to grow, rest, repair damage, fight infection, or stay calm. These cellular conversations form the fundamental communication network that keeps tissues and organs functioning properly. When those signals are distorted, weakened or hijacked, they can drive disease. Their development, called Co-Intelligent Single-cell Spatial Cell-cell Communication (iS2C2), is featured in Signal Transduction and Targeted Therapy, published by Nature.
“Understanding a disease means determining how these cellular conversations have gone wrong and how to repair them,” said Stephen Wong, Ph.D., the John S. Dunn Presidential Distinguished Chair in Biomedical Engineering at Houston Methodist, and the study’s lead author. “In Alzheimer’s disease, for example, disrupted signaling may contribute to inflammation and brain damage. In cancer, abnormal cell-to-cell communication can lead to tumor growth, spread and resistance to treatment.”
To do this, the research team created an AI-enabled technology platform designed to identify, model and explain cellular communication networks in the body. T
“We designed iS2C2 to be both highly accurate and biologically interpretable,” Wong said. “It analyzes complex data, infers how cells may be communicating, determines what causes the disease to develop and explains those findings in simple, biologically meaningful language. This could help researchers move faster from interpreting complex data to gaining a clearer understanding of what is happening and ultimately using this information to develop new treatment recommendations.”
The platform combines advanced mathematical modeling and cell analysis with large-language-model-enabled reasoning, which allows it to also generate more reliable information about how diseased cells behave.
According to Wong, one of the platform’s key advantages is that it can work even when datasets are incomplete. To address limited biological data—a common problem when studying single cells and how they are organized —the platform incorporates generative AI modules to fill missing information, improve predictions and biological interpretation. This helps overcome a major bottleneck in real-world big picture studies.
When applied to Alzheimer’s disease datasets, iS2C2 generated accurate, reproducible and expert-validated results, uncovering previously underappreciated communication pathways in neurons and surrounding support cells in the brain that may contribute to disease progression and could point to new treatment targets.
When applied to bone cancer metastasis data, the platform revealed cell-to-cell communication that contributes to tumor growth in the bone when cancer spreads. Additionally, the technology found a therapy, often used in breast cancer, that could be combined with other treatments to block the spread of bone cancer earlier.
“The fact that this AI platform can point us to a new treatment strategy may be a game‑changer,” Wong said. “If you can identify which cells are driving diseases, how they are communicating and which pathways may be interrupted therapeutically, you create a much more actionable map for precision medicine.”
Wong’s other collaborators on this study are Ju Ahn, Zhihao Wan, Whaohua Qi, Xiaohui Yu, Yuliang Cao, Matthew Vazquez, Amna Irfan, Yuanyuan Zhu, Hong Zhao, Zheng Yin, Alireza Faridar, Lin Wang, Micahel Chan, Daniel Kermany and Wenjuan Dong from Houston Methodist; Zhan Xu, Yunfeng Ding, Fengshuo, Dongxue Mao and Xiang Zhang from Baylor College of Medicine; Doo Yeon Kim from Mass General. Jianting Sheng and Li Yang were the co-corresponding authors.
This study was supported by grants from the National Institutes of Health, the T.T. and W.F. Chao Foundation, Cures Alzheimer’s Fund, the John S. Dunn Research Foundation and the Paul Richard Jeanneret Research Fund.
Journal
Signal Transduction and Targeted Therapy