Why Science? | Stereo-seq, DNA-Nanoball Patterned Arrays and Mouse-Organogenesis Spatial-Transcriptome Evidence

Section 1 of 36

1. Start with a dense coordinate array

Stereo-seq uses patterned arrays of DNA nanoballs carrying spatial barcodes. RNA from a tissue section is captured in situ, sequenced and returned to array coordinates. Dense sampling and a large field can reveal fine patterns across extensive tissue.

Contents · Next section

Section 2 of 36

2. Understand the nanoball role

A DNA nanoball is a compact amplified barcode feature on the array, not a biological ball inside the specimen. Patterned positions create a known coordinate system for captured molecules.

Contents · Previous section · Next section

Section 3 of 36

3. Balance resolution, capture and field of view

The foundational study addressed a common spatial-omics trade-off: fine spatial sampling can shrink the field or reduce sensitivity. Stereo-seq was designed to combine small feature spacing with panoramic coverage. Effective biological resolution still depends on molecules, diffusion and analysis.

Contents · Previous section · Next section

Section 4 of 36

4. Remember the mouse atlas

The 2022 Cell study generated the Mouse Organogenesis Spatiotemporal Transcriptomic Atlas across developing embryos and used spatial patterns to study heterogeneity and cell-fate specification. Its primary evidence is mouse, not human clinical prediction.

Contents · Previous section · Next section

Section 5 of 36

5. Separate feature size from cell resolution

Nanoball spacing can be smaller than a cell, but one feature may capture few molecules. Analysts often aggregate features into bins or segmented cells. The chosen unit changes sensitivity and apparent resolution.

Contents · Previous section · Next section

Section 6 of 36

6. Define the developmental question

Decide whether the aim is organ boundaries, tissue gradients, cell-type organisation or developmental trajectories. State stages, sections and relevant scale. An atlas should answer predeclared questions while preserving room for exploration.

Contents · Previous section · Next section

Section 7 of 36

7. Map the embryo-to-atlas chain

Stage embryos, prepare sections, image morphology, place sections on arrays, capture and barcode RNA, sequence, assign molecules to coordinates, bin or segment data, annotate cell states, align sections and stages, then construct spatial and temporal models.

Contents · Previous section · Next section

Section 8 of 36

8. Stage embryos explicitly

Embryonic day, morphology, litter and collection time can differ. Record them and balance processing. A smooth trajectory cannot repair inaccurate or confounded staging.

Contents · Previous section · Next section

Section 9 of 36

9. Preserve section orientation

A two-dimensional section intersects a three-dimensional embryo. Orientation and anatomical plane determine which structures appear and their apparent distance. Photograph landmarks and use serial sections where needed.

Contents · Previous section · Next section

Section 10 of 36

10. Check array occupancy

Patterned features, barcode quality and background should be measured across the full capture area. Local array failure can resemble an anatomical void. Plot coordinate coverage before gene expression.

Contents · Previous section · Next section

Section 11 of 36

11. Control tissue contact

Wrinkles, incomplete contact or variable permeabilisation change capture. Record tissue boundaries and compare molecule counts with morphology. Large fields make regional quality variation especially important.

Contents · Previous section · Next section

Section 12 of 36

12. Use UMIs and barcode quality

Unique molecular identifiers help count captured molecules, while spatial barcodes assign coordinates. Report valid barcode fraction, mapping, duplicate rate, genes and UMIs across the array.

Contents · Previous section · Next section

Section 13 of 36

13. Choose bin size transparently

Combining neighbouring features improves molecular counts but lowers spatial resolution. Publish the bin dimensions and repeat critical patterns at alternative sizes. A convenient bin should not silently become a biological cell.

Contents · Previous section · Next section

Section 14 of 36

14. Segment cells with morphology

When cell boundaries or nuclei are available, segmentation can aggregate molecules into cells. Embryonic tissues vary in density and shape. Show overlays, missed cells and boundary uncertainty.

Contents · Previous section · Next section

Section 15 of 36

15. Practise with an invented bin table

These fictional bins show why smaller is not always better.

The numbers are teaching examples, not study results.

Contents · Previous section · Next section

Section 16 of 36

16. Inspect known developmental landmarks

Plot well-established tissue markers and compare them with morphology. Agreement supports coordinate and annotation accuracy. Unexpected patterns need independent validation rather than immediate biological storytelling.

Contents · Previous section · Next section

Section 17 of 36

17. Annotate cell states carefully

Reference single-cell RNA-seq can help label spatial profiles. Species, stage, batch and algorithm influence mapping. Report confidence and mixed states; keep inferred labels distinct from directly observed counts.

Contents · Previous section · Next section

Section 18 of 36

18. Build tissue domains

Spatial clustering can identify regions with shared expression. Account for adjacency, sampling density and section effects. Boundaries should be compared across embryos and with anatomy.

Contents · Previous section · Next section

Section 19 of 36

19. Align stages without erasing differences

A spatiotemporal atlas relates embryos collected at different stages. Registration and trajectory models infer correspondence; they do not watch one embryo develop. Show stage-specific data and uncertainty.

Contents · Previous section · Next section

Section 20 of 36

20. Model lineage as a hypothesis

Expression gradients and pseudotime can suggest developmental direction. Destructive sections cannot prove individual-cell ancestry. Lineage tracing, perturbation and live imaging test the hypothesis.

Contents · Previous section · Next section

Section 21 of 36

21. Use embryos as replicates

Nanoballs and bins are nested within sections, sections within embryos and embryos within litters or batches. Independent embryos support developmental generalisation. Dense coordinates do not increase embryo count.

Contents · Previous section · Next section

Section 22 of 36

22. Challenge RNA diffusion

Molecules can spread during permeabilisation and capture. Sharp landmarks, controls and simulations estimate broadening. Nominal nanoball spacing is not the final point-spread function.

Contents · Previous section · Next section

Section 23 of 36

23. Challenge binning choices

Large bins improve sensitivity and smooth the map; small bins reveal detail but increase zeros. Report multiple resolutions and predefine the primary one. Biological conclusions should not flip under modest bin changes.

Contents · Previous section · Next section

Section 24 of 36

24. Challenge section-to-section variation

Adjacent sections are similar but not identical. Tissue loss, angle and morphology change. Use enough sections and specimens to distinguish a repeated developmental feature from one slice.

Contents · Previous section · Next section

Section 25 of 36

25. Challenge atlas completeness

An atlas covers the sampled stages, planes, embryos and detected molecules. Rare transient states can be missed. Name gaps and avoid presenting a reference as a complete organism.

Contents · Previous section · Next section

Section 26 of 36

26. Challenge patent and platform context

The primary paper declared interests related to patents and BGI. Disclosure does not invalidate evidence, but independent replication, open methods and transparent benchmarking remain useful when assessing performance.

Contents · Previous section · Next section

Section 27 of 36

27. Compare spatial owners

Slide-seq uses barcoded bead pucks; DBiT-seq uses crossed microfluidic barcodes; RNA seqFISH+ images sequential optical codes. Stereo-seq owns the patterned DNA-nanoball array and panoramic mouse-organogenesis atlas intent.

Contents · Previous section · Next section

Section 28 of 36

28. Learn with nested grid bins

Students place molecule tokens on a fine grid, then count them using small and large windows. They see sensitivity rise as spatial detail falls. Adding a missing section introduces atlas limitations.

Contents · Previous section · Next section

Section 29 of 36

29. Connect to Primary Science

Young learners can map changes across a developing model or plant diagram and describe where observations occur. The emphasis is ordered evidence, not advanced molecular terms.

Contents · Previous section · Next section

Section 30 of 36

30. Build PSLE Science process skills

Pupils can compare stages, identify variables and controls, graph counts and explain why repeated embryos matter. A fictional atlas encourages precise statements about pattern and uncertainty.

Contents · Previous section · Next section

Section 31 of 36

31. Extend into Secondary and O-Level Science

Biology contributes development and gene expression; Chemistry contributes RNA capture; Mathematics contributes coordinates and scale; Computing contributes segmentation and trajectories. The topic connects syllabus-linked skills to modern research.

Contents · Previous section · Next section

Section 32 of 36

32. Use the topic for school choices

Verify official opportunities in biology, computing, data science, microscopy and ethics. Public atlas data can support authentic inquiry. Never invent programmes, admissions or guaranteed placements.

Contents · Previous section · Next section

Section 33 of 36

33. See the career pathways

Stereo-seq links developmental biology, genomics, tissue preparation, array engineering, statistics and scientific software. Careers require current training and collaborative evidence skills, not an article alone.

Contents · Previous section · Next section

Section 34 of 36

34. Did You Know? One map can hold several useful scales

The same dense coordinate data can be viewed as tiny features, larger bins or segmented cells. Each scale answers a different question and carries a different uncertainty.

Contents · Previous section · Next section

Section 35 of 36

35. Did You Know? An atlas is assembled from snapshots

Each embryo section is fixed at one time. The spatiotemporal story emerges by comparing many snapshots, not by watching the profiled embryo continue developing.

Contents · Previous section · Next section

Section 36 of 36

36. Keep the final claim bounded

A defensible statement names mouse stage, embryo and section counts, array quality, RNA capture, binning or segmentation, annotations, registration and trajectory assumptions. It may report spatial developmental patterns. It should not claim observed ancestry, complete embryos or direct human prediction.

A publication-ready Stereo-seq study begins with a specimen and array ledger. Record embryo, litter, developmental day, morphology, collection time, section plane, array identifier, processing order and operator. Preserve images used for staging and orientation. Without these links, a spatial pattern cannot be separated cleanly from stage, section and batch.

Array and capture quality should be mapped before biological interpretation. Plot valid barcodes, RNA molecules, genes, background and empty-feature fraction across the full field. Large panoramic areas make local failures easy to mistake for anatomical holes. Quality masks should remain visible in every downstream spatial figure.

Binning is an analytical intervention. Pre-register the primary bin or segmentation scale, then repeat critical patterns at finer and coarser sizes. Report how molecule counts, empty rate, boundary sharpness and cell mixing change. A developmental domain that appears only after heavy smoothing is a hypothesis, not a high-resolution observation.

Segmentation and annotation should be evaluated against morphology and independent markers. Show overlays across dense, sparse and oddly shaped tissues. Name the single-cell reference, stage matching and uncertainty. Keep direct gene counts, inferred cell type and inferred lineage as three separate evidence layers.

Spatiotemporal models require embryo-aware validation. Leave out one embryo or stage at a time, compare reconstructed order with known anatomy and test reciprocal or independent datasets when available. Pseudotime and directional fields summarise similarity; they do not observe ancestry. Lineage tracing or perturbation is needed for mechanism.

Statistical inference should preserve coordinates within sections, sections within embryos and embryos within litters or batches. Pre-register primary regions, gradients or transitions and show every embryo-level summary. Millions of nanoball features strengthen spatial sampling within a specimen but do not increase the number of embryos.

Figures should include stage and section diagrams, full-field quality, morphology, alternative bin scales, raw marker maps, segmentation uncertainty, embryo-level summaries and trajectory sensitivity. Disclose platform interests and distinguish replication from demonstration. Panoramic beauty should never conceal uneven capture.

Data stewardship should retain staging images, section maps, array dictionaries, raw sequencing, UMI rules, coordinate matrices, binning and segmentation settings, references, trajectories, code, versions and checksums. The next experiment should address the leading gap: a lineage claim calls for tracing, a boundary claim calls for diffusion calibration, a rare state calls for deeper replicated sampling, and a human implication calls for human evidence. A bounded result remains a mouse atlas finding.

Pre-registration can define embryo stages, primary organs or gradients, array and RNA gates, main bin or segmentation scale, trajectory method and independent embryo unit. Exploratory atlas discovery remains valuable when labelled. This prevents a panoramic dataset from being searched across every scale until one developmental story appears inevitable.

Batch monitoring should use stable RNA or tissue references when feasible. Plot barcode validity, background, RNA molecules, genes, empty features and morphology by array lot and processing date. Stages distributed across runs are far easier to interpret than early embryos on one lot and late embryos on another.

Uncertainty should combine array coordinates, diffusion, section registration, counting, segmentation, stage assignment and embryo variation. Nanoball spacing can be extremely fine while molecule detection and biological boundaries are broader. Report uncertainty at the scale of the actual claim.

Three-dimensional reconstruction from serial sections adds another registration problem. Section loss, compression, angle and developmental asymmetry can break continuity. If a volume is inferred, describe section spacing, alignment landmarks and missing planes. A 3D rendering should not imply a continuously measured embryo.

Cell-fate language needs independent support. Spatial co-location, gene expression and pseudotime can identify candidate progenitors and directions. Lineage tracing, clonal labels or perturbation are needed to show ancestry or necessity. Keep atlas-derived predictions clearly marked.

Negative results require a scale-specific detection boundary. Estimate the smallest domain, rarest state and expression difference supported at the chosen bin and embryo count. A null at a coarse bin may exclude broad regional change while leaving sparse cell populations unresolved.

Protocol transfer requires new validation. Adult organs, tumours, plants and clinical tissue differ from mouse embryos in RNA quality, permeability, cell size and morphology. Re-establish array contact, diffusion, binning and annotation references. Platform density alone does not guarantee equivalent performance.

Reference atlases can become circular if used both to assign and confirm cell types. Validate with withheld markers, independent histology and unsupervised spatial patterns. Report cells or bins that remain ambiguous. A good atlas preserves uncertainty rather than colouring every coordinate.

A concise results sentence might say: ‘Across independently staged mouse embryos, quality-filtered Stereo-seq arrays recovered replicated organ-specific expression domains under a declared binning and registration model.’ It should name embryos, sections and sensitivity. It should not claim observed lineage, complete development or direct prediction for human embryos.

Replicate design should distribute stages, litters and arrays across processing days. Embryos from one litter share genetics and environment, so litter should remain in the hierarchy. Report attempted and included specimens at every stage. A dense atlas from one embryo can be descriptive without representing developmental variability.

Array feature quality and barcode decoding should be monitored with spatial controls. Quantify invalid or low-confidence coordinates and unexpected barcode collisions. Isolated high-expression features may be sequence or contamination artefacts; check raw molecules and neighbouring features before treating them as rare cells.

Integration with single-cell references should avoid data leakage. If the same marker set defines and evaluates a label, agreement is circular. Hold out markers, compare multiple references or validate against morphology. Keep reference-derived probabilities rather than converting every uncertain bin into a hard class.

Spatial statistics should control autocorrelation and multiple testing. Nearby bins are related, and atlas-scale scans create many possible patterns. Predefined developmental regions, corrected exploration and replication across embryos reduce false discovery. A beautiful gradient earns confidence when it repeats independently.

Ethical communication matters whenever developmental findings are discussed. Species and stage should be explicit, and mouse patterns should not be presented as human fetal facts or clinical predictions. Responsible language keeps discovery exciting without outrunning evidence.

Precise verbs keep the record reusable: state whether the study measured, mapped, associated, inferred, predicted or experimentally changed a process. Laboratory work requires approved biological handling, chemical risk controls, trained instrument use and waste disposal; classroom work should use synthetic or public datasets. Final reports should also publish exclusions, failed runs, denominators, effect sizes and uncertainty beside the successful map. That small discipline lets another team reproduce the evidence chain, test a different explanation and decide what the method genuinely adds.

Claims should travel with their scale: molecule, pixel, cell, field, section, specimen or population. Moving between these levels without saying so is a common source of false certainty.

Contents · Previous section · Continue to the Science Learning Hub