Spatially resolved contrastive analysis of perturbation-driven microenvironmental heterogeneity across tissues and conditions
The paper introduces Haruka, a spatially aware contrastive learning framework that disentangles condition-specific microenvironmental changes from shared tissue architecture to systematically analyze perturbation-driven heterogeneity across diverse spatial omics platforms and disease contexts.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Tissues are not merely bags of cells; they are intricate cities where the location of a building determines its function. A cell living in the quiet suburbs of a tissue behaves differently than its identical twin living in the bustling city center, even if they share the same genetic blueprint. For decades, scientists have struggled to map these neighborhoods, often treating every cell as an isolated individual rather than a resident of a specific community. This limitation became a major hurdle when trying to understand how tissues change in response to disease or treatment. When a drug is administered or a virus attacks, the reaction is not uniform; it depends entirely on the local environment surrounding each cell. To truly understand why a treatment works for one person and fails for another, or why a disease progresses in some organs but not others, researchers need a way to separate the permanent architecture of the tissue from the temporary changes caused by the event itself.
A new computational tool called Haruka, developed by researchers at the Whitehead Institute and their colleagues, offers a solution to this problem. Instead of looking at cells in isolation, Haruka acts as a lens that distinguishes between the stable, shared structure of a tissue and the unique, condition-specific changes that occur within it. Imagine trying to hear a specific conversation in a noisy room; you must first understand the background noise to isolate the voice you are interested in. Similarly, Haruka learns the "background" of a tissue—the parts that remain the same whether a patient is healthy or sick, treated or untreated—and then highlights the "salient" parts that change in response to a specific event. By doing this, the tool can reveal hidden patterns of how cells adapt, resist, or fail in their specific neighborhoods, providing a clearer picture of biological complexity than ever before.
The researchers tested this approach first on simulated data and then on real-world biological samples to see if it could outperform existing methods. In these tests, Haruka successfully identified specific regions where cells reacted differently to a perturbation, such as aging or a drug treatment, while correctly ignoring the parts of the tissue that remained unchanged. Unlike previous tools that either ignored the spatial context or failed to separate the signal from the noise, Haruka managed to do both simultaneously. It learned to recognize that a cell's behavior is shaped by its neighbors, and it used this knowledge to map out distinct "microenvironments." These are not just random clusters of cells, but organized communities with shared characteristics that respond to the world in a coordinated way.
One of the most compelling applications of this tool was in the study of melanoma, a type of skin cancer, and how patients respond to immunotherapy. The researchers analyzed tissue samples from six patients before and after treatment, looking for the spatial signatures that predicted whether the therapy would work. Haruka identified four distinct types of immune neighborhoods within the tumors. In patients who responded well to the treatment, the tool found two specific types of neighborhoods where immune cells, particularly a stem-like subset of T cells, were actively engaging with the cancer cells. In one type, these immune cells formed a dense hub of activity, while in the other, they existed in a more balanced partnership with the surrounding tissue structure. Conversely, in patients who did not respond, Haruka revealed two different types of resistant neighborhoods. These areas were characterized by a lack of the crucial immune cells and were dominated by other cell types that created a shield, effectively hiding the tumor from the immune system. The tool showed that the success of the treatment was not just about the presence of immune cells, but about the specific spatial arrangement and the type of neighborhood they inhabited.
The researchers also applied Haruka to human lung tissue to track the progression of fibrosis, a condition where lung tissue becomes stiff and scarred. By comparing healthy lung samples with those from patients suffering from varying degrees of fibrosis, the tool was able to map the transition from health to disease with remarkable precision. It identified specific early warning signs in the cells that were about to become fibrotic, long before the tissue showed obvious signs of damage. This allowed the researchers to see the exact moment when a healthy cell began to change its identity, driven by its neighbors, and start producing the scar tissue that leads to organ failure. The tool proved so effective that it could align tissue samples from different patients with high accuracy, revealing a consistent path of disease progression that was previously difficult to see.
In a study of lung cancer driven by a specific genetic mutation, Haruka helped explain why some tumors become resistant to targeted drugs. The researchers treated mice with a drug designed to kill these cancer cells and then used the tool to examine the surviving tumors. They discovered that the drug resistance was not a random occurrence but was organized into specific spatial niches. In these resistant zones, the cancer cells were surrounded by a unique mix of immune cells and other tissue components that protected them. The tool revealed that these protective neighborhoods were actively suppressing the immune system, creating a safe haven where the cancer could continue to grow despite the presence of the drug. This finding suggests that to overcome resistance, future treatments may need to target not just the cancer cells, but the specific microenvironment that shelters them.
Finally, the researchers pushed the boundaries of the tool by applying it to a complex dataset that measured both gene activity and the accessibility of DNA in the same tissue samples from a mouse model of brain inflammation. This multimodal approach allowed them to see how changes in the genetic code's accessibility led to changes in gene expression within specific brain regions. Haruka successfully separated the shared brain architecture from the specific inflammatory response, identifying distinct regulatory programs that were active in different parts of the brain. When they compared these findings to human genetic data, they found that the specific brain neighborhoods identified by the tool were strongly linked to the genetic risk factors for psychiatric disorders like schizophrenia and bipolar disorder. This connection suggests that the way brain cells organize themselves and respond to inflammation may play a fundamental role in these complex diseases, offering a new way to understand the biological roots of mental illness.
The development of Haruka represents a significant step forward in our ability to understand the spatial logic of life. By separating the constant from the changing, and the shared from the specific, it provides a framework for dissecting the complex interactions that drive health and disease. The tool has already revealed hidden patterns in cancer, fibrosis, and brain disorders, suggesting that the key to many medical mysteries lies not just in the cells themselves, but in the neighborhoods they build. As researchers continue to apply this method to new datasets and diseases, it promises to uncover the subtle, spatially organized mechanisms that govern how our bodies respond to the challenges of the world.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.