| Pre-print |
Platform compared |
Samples |
Gene panels |
Hightlighted data |
|
Systematic benchmarking of high-throughput subcellular spatial transcriptomics platforms
Ren et al., Dec 2024. (1)
|
Xenium
CosMx
|
Treating naive tumors in FFPE from: colon adenocarcinoma (COAD), hepatocellular carcinoma (HCC), and ovarian cancer (OV) |
Xenium: Human 5K
CosMx: Human 6K
Overlapping genes: 2,522
|
Figure 1
Xenium shows greater correlation with scRNA-seq data.
Figure 2
Xenium has lower negative probe signal and higher signal to noise ratio.
Figure 4
Xenium identifies more cell types, has more consistent cell typing across different tools, and has higher correlation with CODEX data for immune cells.
|
|
Comparison of imaging-based single-cell resolution spatial transcriptomics profiling platforms using formalin-fixed, paraffin-embedded tumor samples
Lermi et al., Dec 2024. (2)
|
Xenium
CosMx
Merscope
|
5-µm serial sections of FFPE surgically resected lung adenocarcinoma and pleural mesothelioma samples obtained from 2016 to 2022 and placed in tissue microarrays (TMAs) |
Xenium: Human lung panel (289 genes) + 50 custom genes
Merscope: Human I/O panel (500 genes)
CoxMx: Human 1K
|
Figure 2
Xenium shows lower noise and lower false discovery rates
Figure 5
Manual annotation of the data resulted in Xenium identifying more cell types.
|
|
Single-cell spatial transcriptomics of fixed, paraffin-embedded biopsies reveals colitis-associated cell networks
Mennillo et al., Nov 2024. (3)
|
Xenium
CosMx
Merscope
|
FFPE TMA containing colon biopsies from UC patients and patients without IBD |
Xenium: Custom 290 gene panel
CosMx: Human 1K
Merscope: Custom 280 gene panel
Xenium & CosMx gene overlap: 159
|
Figure 2
Xenium recovers more cell types for both coarse and fine annotations.
Figure 3
Xenium identifies differential gene expression in both the UC and IBD cohorts.
|
|
Comparison of spatial transcriptomics technologies across six cancer types
Cervilla et al., May 2024. (4)
|
Xenium
CosMx
|
6 untreated primary tumor samples, FFPE tissues from colorectal, breast, lung, lymphoma, ovarian, and bladder cancers |
Xenium: Human multi-tissue v1 (377 genes)
CosMx: Human 1K
Overlapping genes: 125
|
Figure 2
When comparing shared scanned areas and genes (125), Xenium detected more cells with a lower false discovery rate.
|
| Single cell approaches define forebrain neural stem cell niches and identify microglial ligands that enhance precursor-mediated remyelination Willis et al., Mar 2024. (5) |
Xenium
Merscope
|
Fresh frozen mouse brain |
Xenium: Custom 347 gene panel
Merscope: Custom 300 gene panel
Overlapping genes: 119
|
Figure 4
Merscope confirmed Xenium’s findings, identifying similar cell types in the same locations..
|
|
Systematic benchmarking of imaging spatial transcriptomics platforms in FFPE tissues
Wang et al., Dec 2023. (6)
|
Xenium
CosMx
Merscope
|
23 FFPE tissue types; 2 TMAs
TMA1 = 7 different cancer types (3–6 patients)
TMA2 = 16 normal tissues (1 patient)
|
Xenium: Human breast, lung, and multi-tissue panels
CosMx: Human 1K
Merscope: Custom panels to match Xenium breast and lung panels
Overlapping genes: 94
|
Figure 2
Xenium has a higher percentage of genes above noise across sample types.
Figure 5
Xenium consistently identified more cell types across samples.
|
|
A comparative analysis of imaging-based spatial transcriptomics platforms
Cook et al., Dec 2023.
|
Xenium
CosMx
|
Prostate cancer FFPE
Serial sections
|
Xenium: Human multi-tissue v1 (377)
CosMx: Human 1K
Overlapping genes: 125
|
Figure 2
Xenium data shows higher correlation with snPATHO-seq data and lower noise.
Figure 4
Xenium had higher sensitivity and dynamic range is consistent in independent samples.
|
Study highlight: How Xenium Prime 5K wins
In a recent preprint, “Systematic benchmarking of high-throughput subcellular spatial transcriptomics platforms,” Ren et al. performed a head-to-head comparison of the Xenium platform with CosMx.
Read on and see for yourself how, in their hands, the Xenium platform:
Identified more cell types
Why it matters: Single cell spatial is fundamentally single cell data; high performance, coupled with the genes included in your assay, should lead to greater ability to discern cell types and better reflect the true biology of your tissue.
Example data:
Takeaway:In the data highlighted above, the Xenium platform identified a greater number of cell types than the CosMx assay it was paired up against. The authors also commented, “Xenium 5K showed a higher proportion of cells consistently annotated as the same cell type across tools, indicating robust annotation reliability,” and that, “Xenium 5K exhibited the most distinct expression patterns, facilitating more accurate cell type annotations."
These findings were consistent with Mennillo et al. 2024, Lermi et al. 2024, and Wang et al. 2023, all of which showed the Xenium platform identifying a greater number of cell types versus CosMx and/or MERFISH.
Better reflected the ground truth of single cell sequencing
Why it matters: Single cell sequencing is the gold standard for single cell transcriptomics, so single cell spatial should correlate with scRNA-seq and snRNA-seq across the entire range of gene expression.
Example data:
Takeaway: Using well-established single cell RNA-seq as a ground truth, the authors demonstrated that the Xenium platform showed higher correlation with scRNA-seq across ~4 logs of dynamic range (Figure 2). In contrast, CosMx showed a lower correlation. Of particular note, these findings were consistent in every study cited above that compared single cell or single nuclei RNA-seq to CosMx, with several researchers suggesting that inflation of low-expression gene counts in CosMx drove the reduced dynamic range.
Offered lower background noise and a higher signal-to-noise ratio
Why it matters:
Accurate single cell spatial imaging data requires genuine transcript calls—not false positives masquerading as data. A high signal-to-noise ratio, coupled with low background noise, helps give you confidence in your assay’s performance.
Example data:
Takeaways: The authors noted that, “After normalization by the total number of signals, Xenium 5K showed a lower ratio of negative control signals compared to CosMx 6K. Notably, negative probes consistently exhibited stronger signals than negative codewords for both platforms, suggesting nonspecific binding to target transcripts are the primary source of background noise in iST platforms.” In the above data, the Xenium platform also exhibited a markedly higher signal-to-noise ratio (Figure 3).
This is important because high target specificity is a requirement for accurate and reliable measurements of transcripts in single cell spatial imaging experiments. High background noise can make it difficult to distinguish low-expression transcripts from background noise (an observation highlighted by both Wang et al. 2023 and Cook et al. 2023) and raise concerns about whether your data is actually data, or false positives.
Forging a brighter spatial future with single cell spatial
Xenium single cell spatial imaging is helping power scientific success by enabling researchers to identify a wide variety of cells, efficiently resolve rare and difficult cell types, and instill confidence with high concordance to single cell sequencing. Read the articles to see for yourself all the ways that the Xenium platform wins versus the competition, and reach out to learn more about what Xenium In Situ can do for your work!
