Imaging spatial transcriptomics is reshaping how scientists study tissues because it reveals not only which genes are active, but also where in the tissue those genes are expressed. This spatial context helps researchers understand how cells communicate, how diseases progress, and how tumors interact with their surrounding microenvironment. A new benchmarking paper led by researchers at the Broad Institute of MIT and Harvard provides one of the most detailed comparisons to date of the leading commercial imaging spatial transcriptomics platforms used for formalin fixed paraffin embedded, FFPE, tissues.

The research team evaluated three major platforms, 10X Xenium, Vizgen MERSCOPE, and Nanostring CosMx, across matched serial sections from tissue microarrays containing 17 tumor types and 16 normal tissue types. These samples represent a wide range of biological diversity and provide a strong foundation for comparing technical performance as well as biological accuracy.

One of the central questions addressed is how effectively each system detects RNA molecules within cells. On genes shared across all platforms, Xenium consistently produced higher numbers of detected transcripts per gene while maintaining strong specificity. This means it captures more signals without introducing many false positives. Xenium and CosMx also showed strong agreement with results from single cell RNA sequencing, demonstrating that both platforms accurately reflect underlying gene expression rather than producing patterns that are artifacts of the imaging process.

Experimental design and iST platforms

a Overall approach for generating iST data. b Different amplification approaches for Xenium, MERSCOPE, and CosMx. c Overview of the tissue types and numbers of cores used in this study. BlC bladder cancer, BrC breast cancer, CRC colorectal cancer, HNSCC head and neck squamous cell carcinoma, Mel Melanoma, NSCLC non-small cell lung cancer, OvC ovarian cancer, SCC squamous cell carcinoma. d DAPI images from the Xenium run of each TMA, including Tumor TMA 1(tTMA1), Tumor TMA 2(tTMA2), and Normal TMA(nTMA).

Cell typing is another critical application of spatial transcriptomics because it enables scientists to identify cell populations and understand how they are arranged in healthy and diseased tissues. All three platforms were able to classify cells into meaningful types, but there were differences in how finely they could subdivide those groups. Xenium and CosMx detected slightly more sub clusters than MERSCOPE, offering a bit more granularity. However, each platform showed different rates of false discoveries and cell segmentation errors, highlighting that higher resolution does not always mean fewer mistakes.

The study also assessed how well each system captured groups of spatially linked genes and recovered gene signatures associated with pathologic features. These capabilities are especially important for cancer research, where understanding the interaction between tumor and immune cells can influence treatment decisions. All three platforms performed this task to varying degrees, giving researchers flexibility depending on whether they prioritize sensitivity, accuracy, or imaging resolution.

Overall, this benchmark serves as a valuable guide for laboratories choosing an imaging spatial transcriptomics platform, especially when working with FFPE samples that are often limited in quantity and critical for diagnosis or retrospective studies. By laying out the strengths and limitations of each technology, the authors provide a resource that helps scientists match the right tool to their research needs. As spatial biology continues to advance, comparisons like this will play an essential role in shaping how the field approaches precision medicine and tissue-based molecular research.