Mantpy: a framework for extracellular matrix analysis in spatial proteomics

In the microscopic world of living tissue, cells do not exist in isolation. They are embedded in a complex, fibrous scaffold known as the extracellular matrix. This material is not merely a passive filler; it is a dynamic structure that provides physical support and chemical signals, guiding how cells behave, move, and communicate. For decades, scientists studying the body's tissues have focused almost exclusively on the cells themselves, treating the surrounding matrix as background noise. This approach has yielded deep insights into cellular function, but it leaves a critical gap in our understanding. Just as a house cannot be fully understood by studying only the people living inside without considering the walls and floors that shape their movement, a tissue cannot be fully understood by looking at cells while ignoring the matrix that holds them together. Modern imaging technologies have finally reached a point where they can capture both the cells and this intricate matrix in their original, undisturbed locations, yet the tools to analyze this combined data have lagged behind.

A new framework called Mantpy addresses this imbalance by offering a way to map the relationship between cells and their surrounding environment as a single, connected system. The researchers behind this work recognized that existing analysis tools were designed with a cell-centric view, forcing scientists to treat the matrix as a separate entity or a static backdrop. Mantpy changes this perspective by representing the tissue as a spatial graph, a digital structure that links points of interest together based on their physical proximity. In this system, the framework builds a network directly from the markers that identify the matrix material, creating a map of the scaffold itself. It then connects this matrix map to a separate map of the cells, weaving them together into a joint network. This allows the software to calculate statistics and apply advanced learning methods that consider how the matrix and cells interact, rather than analyzing them in isolation. The tool is designed to work with data ranging from a single type of matrix marker to complex panels containing many different cellular and matrix signals.

The utility of this approach was demonstrated across several distinct biological settings, revealing details that were previously difficult to see. In human intestine tissue, the framework successfully recovered the layered architecture of the organ, showing how the matrix organizes itself in distinct strata alongside the cells. When applied to infected mouse liver tissue, Mantpy resolved specific changes in the composition and organization of the matrix that were associated with the disease state, highlighting how the scaffold remodels itself in response to infection. Similarly, in mouse lung tissue, the system characterized the specific associations between individual cells and the matrix structures they touched. These findings were not just theoretical; the framework was tested on real, multiplexed datasets that included both cellular and matrix markers, proving its ability to handle complex, multi-layered biological information.

To ensure this tool is accessible to the broader scientific community, the researchers released it alongside the datasets used in their study, which include the extracellular matrix information necessary for others to reproduce the work. The framework is built to interoperate with the scverse ecosystem, a collection of software tools widely used for analyzing single-cell data, ensuring that it fits seamlessly into existing workflows. By extending the unit of analysis beyond the cell to include the matrix that surrounds it, Mantpy provides a more complete picture of tissue biology. It offers a way to see the tissue not as a collection of isolated units, but as an integrated landscape where the scaffold and the cells are inextricably linked, providing a clearer view of how health and disease are shaped by the space between the cells.