Pancreatic cancer is often described as one of the most difficult cancers to detect and treat, but its danger is not explained by malignant cells alone. The disease develops within a densely organized ecosystem of fibroblasts, immune cells, blood vessels, extracellular matrix and nerve-associated structures. These components can surround tumor cells, restrict drug delivery and create local environments that either suppress or support anti-tumor immunity. A new article in Experimental & Molecular Medicine examines how spatially resolved tissue analysis and computational pathology are changing the way scientists interpret this complex architecture.
The article, by SW Bae, A Tsirigos, J Min and colleagues, focuses on a central problem in pancreatic cancer research: conventional pathology often compresses a three-dimensional biological system into a limited number of stained tissue sections. Standard microscopy can reveal the shape and distribution of cells, while molecular assays can identify genes or proteins, but each approach may lose part of the relationship between them. Spatial technologies seek to preserve that relationship by showing not only which molecules are present, but also where they are located within the tumor.
This distinction is crucial in pancreatic ductal adenocarcinoma, the most common form of pancreatic cancer. Tumor cells in this disease are embedded in a prominent desmoplastic stroma, a scar-like tissue compartment rich in collagen, activated fibroblasts and signaling molecules. Rather than acting as passive scaffolding, the stroma can influence tumor growth, immune-cell access and the movement of therapeutic compounds. Spatially resolved methods allow researchers to investigate whether particular cell populations are concentrated at the tumor border, trapped in stromal regions or positioned near blood vessels and ducts.
Among the technologies discussed in this field is spatial transcriptomics, which measures RNA expression while retaining information about tissue coordinates. In a conventional single-cell RNA sequencing experiment, tissue is dissociated into individual cells before analysis. This produces detailed molecular profiles but largely removes the original map. Spatial transcriptomics addresses that limitation by linking gene-expression signals to defined locations on a tissue section. Depending on the platform, the method can survey broad tissue regions or approach cellular resolution, creating molecular maps of tumor and non-tumor compartments.
Imaging-based approaches provide another layer of detail. Multiplex immunofluorescence, imaging mass cytometry and related techniques can detect multiple proteins in the same section, allowing scientists to distinguish cancer cells from immune, stromal and vascular populations. When these measurements are combined with RNA data, researchers can compare cell identity with cell behavior. For example, a region may contain immune cells that are present in large numbers but express molecular features associated with exhaustion or suppression. Such information is more informative than simply counting immune cells across an entire tumor.
The expanding volume of spatial data has made computational pathology an essential part of the process. Digital pathology converts tissue slides into high-resolution images that can be analyzed by algorithms. Machine-learning systems can identify nuclei, classify tissue compartments and quantify features such as cell density, shape, proximity and organization. More advanced models integrate image-derived characteristics with molecular measurements, helping researchers detect patterns that may be difficult to recognize by eye. These analyses can transform a tissue section into a multidimensional map of cellular interactions.
One important goal is to understand the tumor microenvironment as a network rather than a collection of isolated cell types. Computational methods can calculate neighborhood structures, measure distances between populations and identify recurring spatial patterns. A tumor-associated macrophage positioned beside a cancer cell may have a different biological significance from the same macrophage located near a blood vessel or within a collagen-dense stromal region. Spatial analysis can therefore generate hypotheses about how cells communicate through direct contact, soluble factors or physical barriers.
The approach may also improve biomarker development and patient stratification. Pancreatic tumors are biologically diverse, and two patients with similar conventional histological diagnoses may have very different immune landscapes or stromal organization. Spatially defined signatures could eventually help distinguish tumors more likely to respond to immunotherapy, stromal-modifying drugs or combinations involving chemotherapy and targeted treatment. However, the article’s broader message is that such applications require careful validation. Differences in tissue preparation, imaging platforms, computational pipelines and clinical populations can make results difficult to compare.
Several challenges remain before spatial pathology becomes routine in hospitals. Many technologies are expensive, technically demanding and capable of generating datasets that require specialized computational expertise. Tissue sections also provide only a limited view of a tumor that is inherently three-dimensional and heterogeneous. Algorithms must be tested across institutions and patient groups to avoid learning biases related to staining protocols or scanner types rather than genuine biology. In addition, spatial associations do not automatically prove that one cell population causes a particular clinical outcome; experimental and longitudinal studies are needed to establish mechanism.
Nevertheless, the convergence of spatial omics, advanced imaging and artificial intelligence is giving pancreatic cancer research a new visual and molecular language. Instead of asking only which genes are active or how many immune cells are present, investigators can ask how cellular communities are arranged, which barriers separate them and how tissue organization changes during disease progression or treatment. The review by Bae, Tsirigos, Min and colleagues presents this emerging framework as a potential bridge between pathology, molecular biology and precision oncology. By preserving the geography of cancer, spatially resolved analysis may help reveal why pancreatic tumors resist therapy—and where future interventions could be aimed.
Subject of Research: Spatially resolved tissue architecture and computational pathology in pancreatic cancer
Article Title: Spatially resolved tissue architecture and computational pathology in pancreatic cancer
Article References: Bae, SW., Tsirigos, A., Min, J. et al. Spatially resolved tissue architecture and computational pathology in pancreatic cancer. Experimental & Molecular Medicine (2026). https://doi.org/10.1038/s12276-026-01782-4
Image Credits: AI Generated
DOI: 10.1038/s12276-026-01782-4
Keywords: pancreatic cancer, spatial transcriptomics, computational pathology, tissue architecture, tumor microenvironment, digital pathology, spatial omics, precision oncology
Tags: advances in histopathology through computational methodschallenges of 3D tissue analysiscomputational pathology in cancer researchdrug delivery barriers in pancreatic cancerextracellular matrix in pancreatic cancerfibroblast and nerve interactions in tumorsmolecular spatial profiling techniquesPancreatic cancer tissue architecturepancreatic ductal adenocarcinoma tissue mappingspatial genomics in oncologyspatially resolved tumor microenvironment analysistumor immune microenvironment visualization
