Spatial transcriptomics & spatial multi-omics

Spatial transcriptomics measures gene expression while keeping each transcript’s physical position on an intact tissue section, turning a dissociated cell suspension into a coordinate-addressed molecular map. Two complementary chemistry families do this: sequencing-based methods (Visium, Stereo-seq) lay the tissue onto a barcoded array of spots or DNA-nanoball features, capture the released mRNA in situ, then sequence the barcodes back to their coordinates for unbiased whole-transcriptome coverage; imaging-based methods (Xenium, MERSCOPE/MERFISH, CosMx, Molecular Cartography) hybridise fluorescent probe panels to RNA directly on the slide and read out hundreds-to-thousands of targets at subcellular resolution through cyclic microscopy. The global spatial-transcriptomics imager market grew from about 420 million USD in 2023 to a projected 1.28 billion USD in 2026 — a 32.5% CAGR — and by the end of 2025 more than 380 research institutes and pharmaceutical groups had purchased or leased an instrument, with oncology accounting for 54% of revenue and pharmaceutical end-customers rising from 28% in 2023 to 43% in 2026. The technology is shifting from a research curiosity into a drug-discovery and biomarker workflow: a single CosMx Whole-Transcriptome run images over 18,000 human RNA targets per cell in situ, MERSCOPE Ultra maps around 1,000 custom genes across 3 cm² of tissue at ~100 nm pixel resolution, and the same-slide RNA-plus-protein “spatial multi-omics” readout (over 19,000 RNAs with 72 proteins on CosMx Same-Cell Multiomics) is now the competitive frontier.

The key directions of spatial transcriptomics & spatial multi-omics are:

1. NGS-based spatial capture (Visium HD, Stereo-seq): tissue placed on a barcoded spot or DNA-nanoball array for unbiased, whole-transcriptome coverage at multicellular down to subcellular resolution, then sequenced.
2. Imaging-based in situ mapping (Xenium, MERSCOPE, CosMx): cyclic fluorescent probing of targeted 500–1,000+ gene panels at true subcellular resolution, detecting individual transcript molecules.
3. Spatial multi-omics (RNA + protein, same section): CosMx Same-Cell Multiomics, Visium proteins and Akoya PhenoCycler read RNA and 70+ protein markers on one section, linking phenotype to transcriptome in the same cell.
4. AI spatial analysis: deep-learning cell segmentation, spatial-domain identification (Vision-Transformer models) and neighbourhood analysis that turn pixel-level signal into tumour-microenvironment maps and spatial biomarker signatures.

Sectoral value chain#

[Tissue section on spatial slide] ──> [In-situ capture / probe hybridisation] ──> [NGS sequencing OR cyclic imaging]
                                                       │
                                            (barcode → gene decoding)
                                                       │
                                                       ▼
[Spatial biomarker signature] <─── [AI cell segmentation & neighbourhood analysis] <─── [Transcript coordinate map]

Value chain levels#

Cross-cutting technologies of the sector:

• Spatial barcoded arrays: patterned slides — barcoded spots (Visium, ~2 µm for Visium HD) or rolled DNA-nanoball chips (Stereo-seq, features down to ~500 nm) — that tag every captured mRNA with its x/y coordinate before sequencing.
• Error-robust combinatorial barcoding (MERFISH): each gene gets a Hamming-distance-separated binary barcode read over sequential fluorescence rounds, so single-molecule RNA counting stays accurate even with per-round readout errors.
• Deep-learning cell segmentation & spatial AI: convolutional and Vision-Transformer models that assign each transcript to a cell from morphology plus signal, then identify spatial domains and cell-neighbourhood signatures the eye cannot.

The spatial-omics stack couples an in-situ chemistry engine, a high-resolution readout (sequencer or cyclic imager) and a deep-learning analysis layer into one workflow that biopharma buys by instrument, reagent kit or full-service CRO project.

1. In-situ capture and probe chemistry:
   • Sequencing-based platforms pattern barcoded features onto the slide — Visium HD spots down to about 2 µm, Stereo-seq DNA-nanoball features down to roughly 500 nm — so each captured mRNA carries its coordinate before NGS.
   • Imaging-based platforms hybridise oligonucleotide probe panels (MERFISH error-robust barcodes, CosMx direct hybridisation, Xenium in-situ sequencing) that are then read over sequential fluorescence cycles to call single transcript molecules.
2. High-plex readout (NGS + cyclic fluorescence imaging):
   • NGS readout delivers unbiased whole-transcriptome coverage (CosMx ~19,000 genes, Stereo-seq and Visium HD transcriptome-wide), while cyclic imaging reaches 1,000-plus targeted genes at subcellular resolution and 200 nm optical scale.
   • Throughput has risen sharply — CosMx Same-Cell Multiomics now reads over 19,000 RNAs with 72 proteins on one FFPE section, and Akoya PhenoCycler images about 1 million cells in under 10 minutes at 100-plex protein.
3. Spatial bioinformatics and AI:
   • Cell segmentation, spatial-domain detection and neighbourhood analysis run in Seurat, Squidpy and dedicated deep-learning models (Vision-Transformer frameworks) that fuse H&E histology with transcript coordinates to reconstruct single-cell expression.
   • Spatial signatures — the organisation and proximity of cell subsets in the tumour microenvironment — have been shown in a JAMA Oncology meta-analysis of over 8,000 patients to predict anti-PD-1/PD-L1 response more accurately than PD-L1, TMB or GEP alone.