Bruker's Spatial Multiomics update

Bruker CosMx and how it compares to 10X Genomics Xenium

A webinar for Bruker’s Spatial Multiomics Solutions took place recently, and we are going to highlight what’s the state-of-the-art for NanoString’s technologies and new iterations relating to Spatial Biology instruments. A reminder that NanoString was acquired by Bruker about a year or so ago.

When NanoString started in 2008, they introduced nCounter, multiplexed counting of extracted RNA, no amplification, no enzymes involved.

GeoMx launched in 2019 and CosMx a couple of years later. The number of publications and results produced grew up over the year.

Lots of interesting scientific questions being asked using Spatial Multiomics. It is probably one of the most hyped technologies in the last 5 years in biology and biotech.

PaintScape, recently announced at AGBT, allows for 3D Genome analysis, more news on it soon.

GeoMx assays specific areas of your tissue, whereas CosMx is single-cell, even subcellular. CellScape is a robust spatial proteomics platform, to basically validate and optimize the biomarkers discovered with the other platforms. More details on CellScape and how it relates to GeoMx and CosMx at the end of this presentation.

The nCounter technology at the bottom has now evolved into the GeoMx and CosMx equivalents. No amplification, direct targetting of the target mRNAs.

One of the principles behind the design of these platforms is to be highest-plex possible, so that one doesn’t have to choose within a list of genes if the biology is not known.

Focusing on CosMx SMI:

Looking at a human lymph node, every single dot on the right is a single transcript.

CosMx sensitivity: you need to have the best in class segmentation, otherwise, you allocate the transcripts to the wrong cell.

The presentation stressed that high-plexity comes from pure hybridization that doesn’t require amplification (like RCA). The alleged issue with RCA is that it grows into a big blob that takes up more space and reduces sensitivity (this is what Bruker alleges). This is a direct mention to the 10X Genomics Xenium product, which competes in the market with CosMx and uses RCA, via a licensing deal with Harvard University and work done by George Church’s lab many years ago (amongst other IP licenses and technologies involved). It’s also important to note that the paper cited, Deng et al 2018, characterized the brightness and size of RCA amplification products generated with Sequence-Encoded Amplicon (SeqEA) chemistry, which is a distinct type of RCA chemistry that may not be generalizable to all RCA approaches.

The evolution of the platform: recently going from 6k to WTX (19k) coming in 2025. Here is where 10X Genomics Xenium is still behind, currently at 5k with Xenium PRIME, but no news on a whole-transcriptome equivalent to what Bruker has recently announced with the CosMx WTX product.

Segmentation is really important, now in the 5th generation of the cell segmenter, combination of nucleus, cytoplasmic and membrane markers (using Nvidia technology).

Studies comparing different platforms, here one from Northwestern University: 11 patients, 60 blocks, ran CosMx and Xenium side-by-side. The presenter makes the point that “there are a lot of comparisons published out there, but you need to design the experiment properly to compare the technologies in a fair way”.

Results of a comparison of the two platforms (here 6k vs 5k), showing that, in this study, Xenium appears to struggle at high-plex (in the middle). There have been other scientists who have described the pros/cons of 10X Genomics Xenium when it comes to plexity: which one is better? High-plex but fewer unique transcripts/genes per cell, or lower plex but more amplitude in the signal? It’s relevant to point out here that the methods and details in this study have not yet been published or shared in preprint form, and the comparative data has not been made available. There have been several earlier studies comparing CosMx and Xenium (1, 2, 3, 4) including one recent study from Peking University that included a fairly comprehensive direct comparison of Xenium 5K with CosMx 6K assays and the authors have made the data publicly available here. In general, these studies have found that Xenium provided sensitivity that is comparable or higher than CosMx comparisons when considering shared genes, and Xenium has generally provided greater specificity, lower false discovery, and higher concordance with orthogonal benchmarking at the single cell level in these studies. It will be interesting when these data and other independent comparative datasets become available to compare in more detail.

Some examples of using CosMx: pancreatic cancer, macrophages and nerve cells involved in the malignancy.

Here CosMx WTX, unbiased analysis of all the transcripts: inferring the biology from the data. The biology informs the hypothesis.

Breast cancer sample: here examples of 4 pathways, but all of them are covered.

Specific cell types, here CD8 T-cells: why some are in the external stroma domain vs the interior stroma domain?

Again, this highlights the whole transcriptome approach.

Moving to the proteins, the actionable molecule in the cell. Here we look at before and after treatment for the same patient sample.

The final summary of CosMx.

The summary of the biorxiv preprint.

Comparing to 10X Genomics Chromium, which requires more input of biological material. It is of note here that the experimental details and data from this study have not been published or made available to the community, so the comparison is difficult to interpret. Sensitivity of Chromium is dependent on sequencing read depth, which is not specified. Chromium median counts per cell here are quite low compared to the typical range of up to 10,000 counts. It’s also worth mentioning that there is no evaluation shown of the correlation of results, but rather just the total transcripts per cell profiles. This can hide noise in the data, and a simple correlation plot would be a much cleaner way of showing if the CosMx probes are well designed and working. Because of the distributions are notably different, with the CosMx profile more gaussian than the Chromium, it is worth emphasizing that when counting absolute number of mRNA molecules in a cell, the distribution has been shown to be typically right-skewed (long-tailed), as we see here for the Chromium single-cell RNAseq data. We should look more closely at these datasets when available, to understand why the CosMx method is probing different amounts of mRNA molecules than what we would expect from single-cell gene expression datasets.

Higher cells assayed means one can find sparse cell types that wouldn’t be assayed in lower-throughput methods like 10X Genomics Chromium. Below an example of Epsilon cells (130 in a total of 400,000).

Any kind of sample that can be put on a slide is available, e.g. organoids, mixed-cell co-cultures, etc. This is what people call STAMP.

The description of the STAMP approach. This was first published as a way of flowing single cells into a 10X Genomics Xenium slide and then performing the Xenium assay to get a cheaper equivalent than the 10X Genomics Chromium. Since then, 10X announced a newer version of their GEM-X Flex which claims to assay single cells for less than $0.01 per cell.

The multiomics means it’s both proteins and mRNA on CosMx.

Here the protein information gives extra information of the biology.

On the data analysis side, it’s all part of the AtoMx platform.

The CosMx Scratch Space for users to interact with the data examples.

Now looking at specific regions or cell populations within a sample, here with GeoMx.

Here segmenting the tissue to look at alpha cells or beta cells.

GeoMx was initially a proteomics platform. People used it for B-cell discoveries.

GeoMx focus now is on increasing the plexity of the protein panel: 570 proteins in the most recent panel.

Now pushing towards 1250+ proteins for human proteome.

Examples why looking at proteins is important: 570 proteins together with WTA.

Another example: Alzheimer’s disease.

CellScape: validation the biomarkers, here 30 markers.

CellScape is Cyclic IHC.

The EpicIF technology makes the erasing of the fluorescence very efficient and cleaner images.

EpicIF means no photobleaching, no tissue damage.

Pre-developed kits for CellScape or fully customized.

Same-sample re-interrogation, key to make more use of the same sample.

Example of a breast carcinoma, what is the mechanism of immune exclusion? Second assay after that.

The specs of CellScape

Finally, an update on nCounter, since 2008. Launching a multiomics assay.

Same principle, a DNA tag, photocleaved, now combining protein and mRNA.

Here an example of the mRNA, Protein or heatmaps combining the two.

Strategic positioning

Leaving the academic audience aside, there are certain considerations here worth describing for those with an investment interest in 10X Genomics TXG 0.00%↑ and Bruker/NanoString BRKR 0.00%↑.