Xenium, CosMx SMI, and MERSCOPE: Single-Molecule Spatial Platforms Compared

Imaging-based single-molecule spatial transcriptomics is dominated by three platforms: 10x Genomics' Xenium, the CosMx Spatial Molecular Imager (SMI), and MERSCOPE. Each images RNA transcripts directly within intact tissue sections, preserving cell position and morphology, but the underlying chemistries, gene plex, and sensitivity profiles differ in ways that shape which platform suits a given experiment. Understanding those differences before committing to a workflow can save significant time and sample material.

The commercial single-molecule spatial transcriptomics landscape

Single-molecule spatial transcriptomics platforms detect individual RNA transcripts inside a tissue section using fluorescent probes and multiple rounds of imaging, rather than sequencing dissociated cells. This approach retains the spatial coordinates and morphological context of every detected transcript, which is central to studying cell-cell interactions, tissue architecture, and rare cell populations.

Xenium, CosMx SMI, and MERSCOPE all fall into this category, and all three are compatible with formalin-fixed, paraffin-embedded (FFPE) tissue, the preservation method most common in clinical biobanks. Despite this shared goal, the platforms encode and read out transcript identity through different chemistries, which has downstream effects on sensitivity, specificity, and gene panel size. As commercial spatial transcriptomics platforms mature, these three tools sit within a broader landscape of methods and platform choices used across the field, spanning both imaging-based and sequencing-based approaches.

FFPE compatibility matters because it unlocks the vast archives of clinically annotated tissue held in hospital and research biobanks. Fresh frozen tissue remains easier to work with in some respects, but it is far less abundant than FFPE material collected during routine clinical care, so a platform's ability to generate high-quality data from archival blocks directly affects how much existing tissue can be studied retrospectively.

A growing body of independent benchmarking work has compared the three platforms directly, often using tissue microarrays that carry many tumor and normal tissue types on a single slide so that multiple platforms can be tested on serial sections from the same samples. Findings vary by study design, tissue type, and the specific gene panels used, so no single benchmarking paper should be read as a definitive ranking. What follows summarizes how each platform works before turning to a side-by-side comparison of their key specifications.

Xenium's rolling circle amplification chemistry

Xenium detects RNA using padlock probes that hybridize to a target transcript, are enzymatically ligated into a closed circle, and then amplified through rolling circle amplification to generate a bright, localized signal for each transcript. The Xenium instrument then runs multiple cycles of fluorescent probe hybridization and imaging to decode a combinatorial barcode assigned to each gene.

In the original demonstration of the platform, researchers used a 313-gene breast cancer panel imaged at roughly 200 nm per pixel across an area of about 12 mm by 24 mm per slide to resolve tumor heterogeneity and identify rare boundary cell populations at the myoepithelial layer in FFPE breast cancer tissue.

Xenium's standard commercial panels have since expanded to include an option reaching 5,000 genes, alongside the original, smaller targeted panels. Independent benchmarking has also found that Xenium generates higher transcript counts per gene than other platforms on shared genes, without a corresponding loss of specificity. The platform also supports add-on protein subpanels for combined RNA and protein readouts on the same tissue section.

Because the rolling circle amplification step produces a strong, localized signal per transcript, Xenium data tends to show a favorable signal-to-noise ratio even in tissues prone to autofluorescence, such as those containing lipofuscin, elastin, or red blood cells. Turnaround time depends on panel size, with smaller targeted panels completing faster than the largest discovery panels, though the platform generally offers faster sample-to-data timelines than sequencing-based spatial methods.

The CosMx SMI cyclic hybridization platform

CosMx SMI, developed by NanoString and now part of Bruker following its 2024 acquisition of NanoString's spatial biology business, uses direct cyclic hybridization of fluorescently labeled barcode probes rather than enzymatic amplification. Cell segmentation relies on antibody-based morphology stains, which makes the workflow compatible with FFPE tissue.

The platform's original validation study reported high sensitivity (one to two RNA copies detected per cell) and a very low false-call rate, using a combined panel of 980 RNAs and 108 proteins on FFPE lung and breast cancer tissue that identified more than 18 distinct cell types and roughly 100 pairwise ligand-receptor interactions.

CosMx's standard commercial offerings include a targeted panel for broad cell characterization, larger discovery-scale panels, and a protein assay for combined analysis alongside RNA on the same tissue section. In one comparative read-depth study, CosMx SMI yielded the highest number of reads per cell among the platforms compared, reflecting the larger gene panel used in that particular study rather than a fixed advantage in per-gene sensitivity.

Since the Bruker acquisition, CosMx has continued to expand its panel catalog, including whole-transcriptome and larger discovery-scale offerings alongside its original targeted panels. The platform's morphology-based segmentation, which uses antibody stains rather than relying solely on nuclear signal, is designed to help distinguish closely packed cells in dense tissue regions such as tumor-immune interfaces.

MERSCOPE: a combinatorial MERFISH platform

MERSCOPE commercializes multiplexed error-robust fluorescence in situ hybridization (MERFISH), a technique that labels RNA transcripts with combinatorial, error-correcting binary barcodes and reads them out across sequential rounds of two-color imaging. The original method paper describing this combinatorial barcoding approach demonstrated simultaneous imaging of 100 to 1,000 RNA species in individual cells, using encoding schemes designed to detect and, in some cases, correct single-molecule labeling errors.

Vizgen's commercial MERSCOPE panels are typically built around targeted, several-hundred-gene configurations, such as the MERSCOPE Immuno-Oncology Panel used in several published platform comparisons, with options to extend panel content for specific research questions. Because MERFISH barcodes are error-robust by design, the platform can maintain reliable transcript identification even as panel size grows.

Across published comparisons, MERSCOPE has generally shown lower per-gene sensitivity than Xenium on shared gene panels in some benchmarking datasets, though the gap narrows or shifts depending on tissue type, panel content, and the specific run conditions used in a given study.

MERSCOPE's imaging workflow and desktop visualization software are designed to output cell segmentation results, along with raw imaging data, in formats compatible with common open-source analysis packages. This has made the platform a frequent choice in published studies of tumor immunology and tissue architecture where researchers want direct access to raw imaging files alongside processed transcript tables.

Comparing plex, resolution, and sensitivity across single-molecule spatial platforms

No single benchmarking study has produced a universal ranking of the three platforms, largely because each comparison uses different tissue types, gene panels, and experimental conditions. Even so, several consistent patterns emerge across the published literature.

Xenium has repeatedly shown strong per-gene sensitivity and specificity on genes shared across platforms in independent comparisons, while CosMx SMI often yields higher total reads per cell when its larger panels are used. MERSCOPE's combinatorial MERFISH chemistry offers strong error correction, which supports reliable calls even as researchers scale up panel size for a given experiment.

Beyond raw sensitivity, published comparisons have also looked at how well each platform supports downstream cell typing. Several studies have found that all three platforms can perform spatially resolved cell typing with reasonable accuracy relative to matched single-cell reference data. Xenium and CosMx SMI have sometimes resolved more distinct cell subclusters than MERSCOPE on the same tissue, likely reflecting differences in capture efficiency for weakly expressed marker genes. Segmentation quality, which depends on the nuclear or morphology stain used by each platform, also affects how cleanly transcripts are assigned to individual cells in densely packed tissue.

Throughput and cost also factor into platform choice, though neither is captured well by plex or sensitivity figures alone. Larger discovery panels generally require longer imaging cycles and produce larger raw data files, which has downstream implications for storage and analysis infrastructure regardless of which of the three platforms is used. Core facilities weighing platform adoption often base their decision on which chemistry best matches the tissue types and questions most common in their user base, rather than aiming to support every possible panel size or plex option.

In one head-to-head lung tumor comparison, researchers ran a roughly 1,000-plex CosMx Universal Cell Characterization Panel, a 500-plex MERSCOPE Immuno-Oncology Panel, and a 289-plex Xenium lung panel with 50 custom add-on genes side by side on the same tissue microarray, illustrating the range of standard panel sizes used across the three platforms in practice.

Table 1: A comparison of the standard specifications and chemistries for Xenium, CosMx SMI, and MERSCOPE.

Feature	Xenium	CosMx SMI	MERSCOPE
Vendor	10x Genomics	Bruker (formerly NanoString)	Vizgen
Detection chemistry	Padlock probe ligation with rolling circle amplification	Direct cyclic hybridization of barcode probes	Combinatorial MERFISH with error-robust binary barcoding
Typical standard panel	Approximately 300-plex to 5,000-plex	Approximately 1,000-plex, with larger discovery panels available	Approximately 500-plex, with extended options available
Protein add-on	Yes	Yes	Limited
FFPE compatibility	Yes	Yes	Yes

Choosing among the three platforms typically comes down to a short set of practical questions:

1. Does the project need a targeted panel of a few hundred genes, or a larger discovery-scale panel of several thousand?
2. Is combined RNA and protein detection on the same section required?
3. How does each vendor's tissue preparation and imaging workflow fit existing core facility infrastructure?
4. Have independent benchmarking studies been published on a comparable tissue type to the one under study?

Framing the decision this way, rather than around a single headline specification, tends to produce a better match between platform and research question.

Choosing the right platform: Xenium, CosMx SMI, or MERSCOPE

Xenium, CosMx SMI, and MERSCOPE each provide a viable route to single-molecule, subcellular spatial transcriptomics, but the right choice depends on the specific combination of panel size, sensitivity requirements, and multiomic needs a project demands. Independent benchmarking data can help narrow the decision, but results should be weighed against the tissue type and panel used in each study. A platform's relative performance can shift meaningfully between, for example, a densely cellular lymphoid tissue and a more sparsely populated stromal section.

As all three platforms continue to expand their panel offerings and refine their chemistries, researchers evaluating single-molecule spatial platforms are well served by revisiting the published comparisons closest to their own tissue type before committing to a workflow. These platforms represent just one piece of a much larger set of methods and applications shaping how researchers study tissue in its native context.

This content includes text that has been created with the assistance of generative AI and has undergone editorial review before publishing. Technology Networks' AI policy can be found here.