Biology & Life Sciences

RAEFISH: Reading Every Gene's Location Without Sequencing a Single Base

Spatial transcriptomics has long required sequencing to read gene expression in tissue. RAEFISH demonstrates that imaging alone can detect all 23,000 human genes at single-molecule resolutionโ€”no sequencer needed.

By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.

Spatial transcriptomics won Nature Methods' Method of the Year in 2020, and for good reason: knowing where genes are expressed within a tissueโ€”not just whether they are expressedโ€”transforms our understanding of development, disease, and cellular organization. But every major spatial transcriptomics platform to date has depended on sequencing at some stage of the workflow. Sequencing introduces cost, latency, and infrastructure requirements that constrain who can perform these experiments and how quickly results arrive.

RAEFISH takes a different approach entirely. Rather than sequencing transcripts after capturing them in situ, RAEFISH reads gene identity and location through imaging alone, achieving whole-genome spatial transcriptomics at single-molecule resolution without a sequencer ever touching the sample.

The Landscape: How Spatial Transcriptomics Works Today

Current spatial transcriptomics methods fall into two broad families:

Sequencing-based approaches (e.g., 10x Visium, Slide-seq) capture mRNA from tissue sections onto barcoded arrays, then sequence the captured molecules to determine both gene identity and spatial coordinates. These methods offer genome-wide coverage but at limited spatial resolutionโ€”typically 10โ€“55 micrometers per spot, meaning each measurement averages across multiple cells.

Imaging-based approaches (e.g., MERFISH, seqFISH+) use combinatorial fluorescence in situ hybridization to detect individual RNA molecules directly in tissue. These achieve single-molecule resolution but have historically been limited to pre-selected gene panelsโ€”hundreds to a few thousand genes, far short of the full transcriptome.

The trade-off has been clear: sequencing-based methods see the whole genome at coarse resolution; imaging-based methods see selected genes at fine resolution. Neither has delivered both genome-wide coverage and single-molecule spatial precision in a sequencing-free workflow.

What RAEFISH Does Differently

RAEFISH bridges this gap. According to the published work, the method achieves detection of approximately 23,000 human genes at single-molecule resolution through an imaging-only pipeline. No library preparation, no flow cells, no sequencing runs.

The core innovation lies in how gene identity is encoded and decoded through sequential rounds of hybridization and imaging. While previous imaging-based methods required selecting a panel of target genes in advanceโ€”constrained by the combinatorial encoding capacity of their probe setsโ€”RAEFISH expands this capacity to cover the entire coding transcriptome.

The authors report improvements in cost, speed, and resolution compared to existing sequencing-dependent approaches such as 10x Visium. By eliminating the sequencing step entirely, the workflow compresses what has traditionally been a multi-day process (tissue preparation โ†’ capture โ†’ library prep โ†’ sequencing โ†’ data processing) into an imaging-centered pipeline.

Claims and Evidence

<
ClaimSourceVerdict
RAEFISH achieves sequencing-free whole-genome spatial transcriptomicsCell, 2025 โ€” core method demonstrationStated in abstract
Covers approximately 23,000 human genesCell, 2025 โ€” probe design targeting full coding transcriptomeStated in abstract
Operates at single-molecule resolutionCell, 2025 โ€” individual transcript detectionStated in abstract
Improved cost, speed, and resolution vs. existing approaches (e.g., 10x Visium)Cell, 2025 โ€” comparative benchmarkingStated in abstract; magnitude of improvement requires full-text verification

Critical Analysis

Several aspects warrant careful examination before assessing RAEFISH's practical impact.

Sensitivity and false discovery rates. Detecting 23,000 genes through imaging raises questions about optical crowdingโ€”when transcript density is high, distinguishing overlapping signals becomes increasingly difficult. The abstract does not specify false positive or false negative rates, which are critical for downstream biological interpretation. A method that detects all genes but misassigns 5% of molecules to the wrong gene would produce systematically misleading spatial maps.

Tissue compatibility. Spatial transcriptomics methods vary in their compatibility with different tissue types. Thick tissues, lipid-rich tissues (brain), and highly autofluorescent tissues (liver, kidney) each present distinct optical challenges. Whether RAEFISH maintains its resolution claims across diverse tissue types remains to be established.

Throughput and scalability. Imaging-based methods require sequential rounds of hybridization, each adding hours to the experimental timeline. For 23,000 genes, the number of imaging rounds neededโ€”and the associated photobleaching, tissue degradation, and registration errorsโ€”could be substantial. The abstract claims improved speed, but the absolute time per sample matters for adoption.

Computational demands. Decoding single-molecule identities from sequential imaging rounds across an entire tissue section generates large image datasets requiring specialized analysis pipelines. The computational infrastructure needed may offset some of the cost advantages gained by eliminating sequencing.

Open Questions

  • What are the sensitivity and specificity benchmarks? How does RAEFISH's per-gene detection sensitivity compare to bulk RNA-seq or existing imaging methods like MERFISH when both are applied to the same tissue?
  • How does the method scale to larger tissue areas? Single fields of view at high magnification cover small areas. Tiling entire tissue sections introduces registration challenges that compound with each imaging round.
  • Can RAEFISH capture transcript isoforms? Gene-level detection is valuable, but many biological questions require isoform-level resolution. Whether the probe design supports splice variant discrimination would significantly affect the method's utility.
  • What is the practical cost comparison? "Improved cost" relative to 10x Visium needs quantification. Visium's per-sample cost has decreased substantially as the platform has matured, and a meaningful comparison requires accounting for equipment amortization, reagent costs, and labor time.
  • Will this enable clinical applications? Sequencing-free workflows are attractive for pathology settings where turnaround time matters. But clinical adoption requires regulatory validation and standardized quality metrics that are not yet established.
  • Closing Reflection

    RAEFISH represents a notable technical achievement: demonstrating that the entire human coding transcriptome can be spatially resolved through imaging without sequencing. If the sensitivity and specificity metrics hold up under independent replication, the method could substantially lower the barrier to entry for spatial transcriptomicsโ€”particularly in laboratories and clinical settings that lack sequencing infrastructure. The critical question is not whether the method works in principle, but whether it works robustly enough, across enough tissue types, to displace the sequencing-based workflows that have become the field's default.

    References (2)

    RAEFISH achieves sequencing-free whole-genome spatial transcriptomics at single-molecule resolution. Cell (2025). DOI: 10.1016/j.cell.2025.01.037.
    Ou, X., Ma, C., Sun, D., Xu, J., Wang, Y., Wu, X., et al. (2025). SecY translocon chaperones protein folding during membrane protein insertion. Cell, 188(7), 1912-1924.e13.

    Explore this topic deeper

    Search 290M+ papers, detect research gaps, and find what hasn't been studied yet.

    Click to remove unwanted keywords

    Search 6 keywords โ†’