First-Of-Its-Kind AI Tool Visualizes Cell’s ‘Social Network’ To Treat Cancer

By LabMedica International staff writers
Posted on 18 Mar 2025

Image: The generative AI tool can interpret millions of cells in human tissues in hours (Photo courtesy of Wellcome Sanger Institute)

Every cell in the human body is engaged in constant communication with its environment and operates within a broader network of interactions. Each cell has specific features that help identify it as part of its network, such as the proteins expressed on its surface. These features allow similar cells to be connected based on their characteristics. Advances in single-cell and spatial genomic technologies have greatly enhanced our understanding of the human body, leading to the creation of comprehensive cell atlases that map different tissues and organs. These atlases provide valuable information on various cell types, their locations, and how genetic variations affect their interactions. By understanding the cellular mechanisms of the human body, researchers can gain insights into diseases and identify potential drug development targets. While these atlases offer detailed information about the locations of cells and how they interact within their specific microenvironments, quantifying and interpreting these cellular neighborhoods, as well as understanding the driving forces behind their social interactions, remains a challenge. Now, an innovative AI-based tool is available to interpret millions of human tissue cells in just hours, offering new insights and enabling researchers and clinicians to explore conditions like cancer.

This first-of-its-kind AI-based neural network created by researchers from the Wellcome Sanger Institute (Cambridgeshire, UK) and their collaborators can rapidly analyze and interpret millions of cells from a patient sample, predicting molecular changes in the tissue. It holds the potential to identify where personalized treatments might be most effective for conditions like cancer. The groundbreaking tool, called NicheCompass, was introduced in a paper published in Nature Genetics and utilizes generative AI to create a visual database that integrates spatial genomic data on cell types, their locations, and their interactions. This is part of the broader Human Cell Atlas Initiative. NicheCompass is the first AI-based method capable of quantifying and interpreting a range of data, from cell-to-cell communication to identifying and analyzing different cellular neighborhoods.

NicheCompass operates on a deep-learning AI model based on cell-to-cell communication. It learns how cells communicate within their networks and then categorizes these interactions, grouping cells into neighborhoods within tissues based on shared features. This allows NicheCompass to interpret the data, enabling researchers and clinicians to ask critical questions and gain a deeper understanding of health conditions. For instance, NicheCompass has been used to uncover tissue changes in patients with breast and lung cancer. Ultimately, the AI tool will facilitate the development of personalized treatment plans, pinpointing specific molecular alterations that could be targeted in cancer therapies. Researchers have demonstrated that NicheCompass can identify how different individuals may respond to treatments—within just one hour—using the power of AI. In a study, data from 10 lung cancer patients was analyzed, and similarities and differences between the patients were revealed.

These similarities enhance the general understanding of cancer and highlight transcriptional changes that could inform the development of new treatments. On the other hand, the differences point to new potential avenues for personalized therapies. As more patient data becomes available, clinicians will be able to input their patient’s data into NicheCompass and receive a detailed analysis within an hour, aiding clinical decision-making. The team also applied NicheCompass to breast cancer tissue, showing that it can be used effectively across different types of cancer. In another experiment, NicheCompass was used on a spatial atlas of a mouse brain containing 8.4 million cells, successfully identifying brain regions and generating a visual map of the entire organ. This illustrates the tool’s potential to analyze spatial atlases of entire organs, a resource being developed by researchers globally.

“Having a huge amount of data about the human body is crucial to finding new ways to understand, prevent and treat disease,” said Sebastian Birk, first author at the Institute of AI for Health, Helmholtz Munich and the Wellcome Sanger Institute. “However, we also need tools that allow us to access all the benefits this information could provide. NicheCompass is a significant leap in this field, leveraging the power of AI but also offering interpretability, allowing researchers and clinicians to ask questions about their data and better understand and treat diseases.”

Related Links:
Wellcome Sanger Institute