Imaging-Based Spatial Transcriptomics: MERFISH, seqFISH, and In Situ Sequencing

MERFISH (multiplexed error-robust fluorescence in situ hybridization), seqFISH (sequential fluorescence in situ hybridization), and in situ sequencing (ISS) place individual RNA transcripts on a map of the cell at subcellular resolution, something bulk and even single-cell sequencing cannot do. These imaging-based spatial transcriptomics methods trade whole-transcriptome coverage for precision, reading out a curated gene panel through repeated rounds of hybridization or sequencing directly on intact tissue. Rather than dissociating a sample into single cells and losing spatial context, each method images RNA in place, then uses multiple rounds of signal readout to identify far more genes than a single round of fluorescent staining could.

How multiplexed FISH enables imaging-based spatial transcriptomics

Multiplexed fluorescence in situ hybridization (FISH) enables imaging-based spatial transcriptomics by encoding each RNA species as a unique barcode read out across repeated rounds of hybridization and imaging, rather than assigning one color per gene. Standard single-molecule FISH labels only one or a few RNA species per imaging round, so scaling to hundreds or thousands of genes by microscopy alone once seemed impractical. The original demonstration of this approach showed that combinatorial labeling with error-robust encoding schemes could image 100 to 1,000 distinct RNA species within individual cells.

This barcoding logic separates imaging-based spatial transcriptomics from single-round fluorescent staining, letting a handful of color channels distinguish a much larger gene panel. Error-robust encoding adds a further safeguard, building redundancy into each barcode so a single misread round does not misassign a transcript. The result is a family of related methods, including MERFISH, seqFISH, and ISS, that share this core barcoding logic while differing in probe chemistry and readout strategy. For researchers comparing platforms, this shared foundation is why panel design, not fluorophore count, is usually the more consequential variable, a point examined later in this guide. Together, these approaches make up one branch of the broader spatial transcriptomics methods and platforms landscape, complementing sequencing-based techniques such as Visium, Slide-seq, and Stereo-seq that capture transcriptome-wide data from barcoded capture arrays.

Decoding these barcodes is a computational task as much as an imaging one. Each round of the experiment produces a new image, and software must align successive images, call spots, and translate each spot's multi-round signal pattern back into a gene identity before any biological analysis can begin. Cell segmentation and spot-calling software must keep pace with imaging throughput, since a backlog in image processing can offset any gains from faster imaging cycles. Error-robust codes are designed with enough separation between valid barcodes that a single dropped or extra signal in one round still resolves to the correct gene, rather than being discarded or misassigned.

How MERFISH works

MERFISH assigns each RNA species a binary barcode built from error-robust codewords, then reads that barcode across repeated rounds of single-molecule FISH to determine a transcript's identity and position within the cell. Later refinements to the original approach increased measurement throughput by roughly two orders of magnitude, enabling gene expression profiling of tens of thousands of cells within an 18-hour measurement.

Because MERFISH relies on direct imaging rather than sequencing, its practical panel size is constrained by how many barcodes can be reliably distinguished within the microscope's available color channels and imaging rounds. Most MERFISH experiments target curated panels of several hundred to a few thousand genes rather than the whole transcriptome, a trade-off examined later in this guide. Commercial implementations of the chemistry, including Vizgen's MERSCOPE platform, have made the approach accessible to laboratories without custom optics or in-house imaging pipelines.

Raw MERFISH output is a set of decoded transcript positions across a tissue section, not yet a cell-by-cell expression matrix. Turning that output into usable data requires cell segmentation to draw boundaries around individual cells, followed by assignment of each decoded transcript to the cell it falls within. Segmentation accuracy directly affects data quality, since poorly drawn cell boundaries in densely packed tissue can misassign transcripts between neighboring cells before any downstream analysis occurs. Because MERFISH data are inherently sparse compared with bulk sequencing, researchers typically pool signal across multiple cells or apply specialized statistical methods when working with lower-abundance transcripts.

How seqFISH and seqFISH+ work

seqFISH barcodes transcripts through successive rounds of hybridization, imaging, and probe stripping, reading a fluorophore sequence rather than a single static color to identify each RNA species. The original sequential barcoding scheme demonstrated this approach in single cells using a small set of fluorophores across multiple hybridization rounds.

seqFISH+ extended this design by combining pseudocolor barcoding with additional hybridization rounds, pushing gene counts toward transcriptome scale while maintaining sub-diffraction-limit resolution. In one demonstration, seqFISH+ profiled 10,000 genes in single cells within mouse brain tissue, distinguishing cell classes and revealing subcellular RNA localization patterns and ligand-receptor pairs between neighboring cells. That study imaged the cortex, subventricular zone, and olfactory bulb using a standard confocal microscope, showing that transcriptome-scale imaging did not require specialized optical hardware beyond what many core imaging facilities already operate.

The core distinction between MERFISH and seqFISH lies in how each barcode is built. MERFISH encodes genes as binary, error-robust codewords read out by the presence or absence of signal in each round, while seqFISH and seqFISH+ encode genes as ordered sequences of colors, or pseudocolors, across rounds. Both strategies distinguish many genes using few fluorophores across many imaging cycles, but their error-correction properties and achievable panel sizes differ.

How ISS works

ISS reads short RNA sequences directly within preserved tissue using padlock probes that circularize on a target transcript, followed by RCA and sequencing by ligation performed under the microscope. The original demonstration of this approach identified point mutations and gene expression directly in tissue sections without dissociating cells, preserving the histological context that homogenization-based methods lose.

Commercial developers later adapted this padlock-probe chemistry into single-molecule imaging platforms such as 10x Genomics' Xenium, which pairs targeted in situ gene panels with automated imaging and analysis software. This lineage connects the earliest in situ sequencing demonstrations to the high-throughput commercial platforms now used across many research and clinical laboratories.

Because ISS can detect specific sequence variants rather than only gene identity, it has found particular use in cancer research, where distinguishing a mutant transcript from its wild-type counterpart in its original tissue location can reveal how mutation status varies across a tumor. The original demonstration of this capability detected rare mutant KRAS transcripts within a mixed cell population, showing that ISS can localize a specific mutation to individual cells rather than only reporting its overall frequency across a sample. Most other imaging-based methods, which typically detect gene identity rather than sequence variants, do not offer this capability directly.

Comparing resolution, panel size, and throughput

Because imaging-based methods read out a fixed panel of barcoded genes, panel design, not sequencing depth, determines a study's biological scope. Comparative benchmarking across commercial single-molecule platforms, including CosMx, MERFISH, and Xenium, has found that resolution and sensitivity vary meaningfully by tissue type and panel design, even among platforms built on similar barcoding logic.

That same benchmarking study found that platform performance depended on more than resolution alone. Tissue age affected data quality unevenly. Two of the three platforms lost signal strength in older archival samples, while the third performed consistently regardless of tissue age. Negative control design also differed across platforms, and one panel lacked a dedicated negative control probe set entirely, which limited how directly its background signal could be benchmarked against the other two. Findings like these illustrate why a platform that performs well in one published comparison will not necessarily generalize to a different tissue type, sample age, or research question.

The trade-off is consistent across the field. Smaller, curated panels typically achieve higher per-gene sensitivity and faster imaging, while larger panels approach whole-transcriptome coverage at the cost of additional imaging rounds and processing time. Studies pairing single-cell, spatial, and in situ platforms on the same tissue section, including work combining Xenium with single-cell and spatial sequencing in breast cancer tissue sections, have used this complementarity to resolve rare cell populations at tissue boundaries.

Optimize Your mRNA Testing Workflows

This brochure outlines a comprehensive suite of analytical services aligned with the latest guidelines and standards from the United States Pharmacopeia (USP).
View How To Guide

Advertisement

Table 1: A qualitative comparison of imaging-based spatial transcriptomics methods by readout strategy, resolution, and typical panel scope.

A practical framework for choosing among these methods can help narrow the decision before committing imaging time and reagent cost to a full-scale study.

1. Define the required gene panel, including whether the study needs a small, hypothesis-driven marker set or broader, transcriptome-scale coverage.
2. Match resolution requirements to the biological question, since subcellular RNA localization studies demand different validation than cell-typing studies.
3. Confirm tissue compatibility, particularly for formalin-fixed, paraffin-embedded archival samples versus fresh frozen sections.
4. Assess available imaging infrastructure, since panel size and hybridization rounds both extend total run time.
5. Pilot the candidate gene panel on representative tissue before committing to a full-scale experiment.

Reagent cost and imaging time both rise with panel size and hybridization rounds, so a larger gene panel is rarely a simple upgrade. Laboratories weighing an imaging-based platform against a sequencing-based one typically trade a smaller, precisely localized gene panel and longer imaging time for the subcellular resolution that whole-transcriptome sequencing-based methods cannot match.

Choosing between MERFISH, seqFISH, and in situ sequencing

MERFISH, seqFISH, and in situ sequencing each convert a curated set of transcripts into a spatial map of the cell, trading transcriptome-wide coverage for the resolution and positional detail that dissociation-based methods cannot provide. Selecting among them depends less on which method is theoretically capable of measuring more genes and more on the specific resolution, throughput, and tissue constraints of a given experiment. A hypothesis-driven study built around a short, well-validated gene list is usually well served by MERFISH or seqFISH, while a study centered on point mutations or other sequence-level differences is better matched to ISS.

As commercial implementations of these chemistries mature, the practical differences between platforms increasingly come down to panel design, imaging infrastructure, and analysis support rather than the underlying barcoding principle. Within the wider spatial biology methods and technologies landscape, these imaging-based approaches represent the highest-resolution branch of spatial transcriptomics available today. That advantage narrows for researchers who need broad, hypothesis-free gene discovery, where a sequencing-based platform remains the more suitable starting point. Researchers evaluating an imaging-based platform benefit from testing a candidate gene panel against their own tissue type before committing to a full study.

None of the three methods is a universal replacement for the others, and many research programs end up using more than one across different stages of a project. A small, hypothesis-driven MERFISH or ISS panel can validate candidate markers identified through unbiased sequencing-based spatial transcriptomics or single-cell RNA sequencing, closing the loop between discovery and spatial confirmation within one workflow. As panel sizes, decoding software, and commercial support continue to improve across all three chemistries, the practical gap between them is likely to keep narrowing.

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.