Why Science? | CytoSignal, Spatial Ligand–Receptor Scores and Signalling-Dynamics Evidence

Section 1 of 36

1. Start with a location-and-change question

CytoSignal asks two linked questions: where does a tissue contain RNA evidence compatible with a ligand–receptor interaction, and how might that evidence be changing? The first produces a spatial score; the VeloCytoSignal extension estimates a derivative of that score from RNA velocity. Neither quantity directly observes molecules crossing space, binding a receptor or activating a pathway.

Contents · Next section

Section 2 of 36

2. Understand the static score

For an interaction, the framework combines ligand information around a receiving location with receptor expression at that location. The 2026 paper describes an LRscore that can be written conceptually as ligand support multiplied by receptor support. The biological meaning depends on the neighbourhood rule, transcript detection and interaction catalogue, so a bright pixel is evidence under a model rather than a photographed event.

Contents · Previous section · Next section

Section 3 of 36

3. Separate contact from diffusion

CytoSignal treats contact-dependent and diffusion-dependent interactions differently. Contact biology calls for a tight local neighbourhood; a diffusible ligand can draw information from a wider field. This distinction is scientifically helpful because it makes range an explicit assumption. It also means that assigning the wrong interaction class can move, blur or manufacture an apparent signalling hotspot.

Contents · Previous section · Next section

Section 4 of 36

4. See what VeloCytoSignal adds

RNA velocity estimates a local direction of transcriptomic change from spliced and unspliced RNA. VeloCytoSignal applies velocity information to ligand and receptor components to estimate the time derivative of the LRscore. A positive derivative suggests increasing RNA-based compatibility under the model. It does not promise that future protein abundance or signalling flux will rise.

Contents · Previous section · Next section

Section 5 of 36

5. Look for spatial gradients

A tissue may show a graded ligand field rather than a sharp sender–receiver border. CytoSignal can make such patterns visible at cellular resolution when the input technology and segmentation support that claim. Interpret the gradient against anatomy, cell density, capture depth and tissue edges. A smooth computational surface can also arise from the smoothing rule itself.

Contents · Previous section · Next section

Section 6 of 36

6. Connect interactions to genes

The framework can identify genes associated with local signalling scores, helping researchers move from a map of candidate interactions to possible cellular responses. Association is a prioritisation step. Shared cell identity, stress, location or technical quality can influence both the score and gene expression, so associated genes should be tested with covariates, independent specimens and perturbation.

Contents · Previous section · Next section

Section 7 of 36

7. Keep the publication boundary

The peer-reviewed Nature Genetics article was published in 2026 and evaluated the method on studied spatial-transcriptomic systems, with in situ proximity ligation assay validation in reported settings. Those results support the framework’s evaluated claims. They do not validate every inferred interaction in a new organ, species, disease, platform or spatial resolution.

Contents · Previous section · Next section

Section 8 of 36

8. Build a spatial specimen manifest

Record organism, organ, anatomical region, condition, donor, section, assay, capture unit, segmentation method, coordinates and preprocessing. State whether a row represents a cell, bead or mixed spot. The phrase ‘cellular resolution’ should be reserved for inputs that genuinely resolve cells; software cannot restore identities that the measurement never separated.

Contents · Previous section · Next section

Section 9 of 36

9. Preserve coordinate provenance

Archive the original coordinate system, scaling, rotations, image registration and any excluded regions. A flipped axis may not alter a heat map’s beauty but can reverse anatomical interpretation. Keep tissue masks and boundary annotations beside the expression matrix so every neighbourhood can be reconstructed and inspected.

Contents · Previous section · Next section

Section 10 of 36

10. Curate the interaction catalogue

Version ligand–receptor identities, species mappings, complex rules and contact-versus-diffusion labels. A receptor complex requires the relevant subunits; a ligand family may include context-dependent relationships. Publish excluded interactions and reasons. The candidate universe is part of the evidence because changing it changes both discoveries and the multiple-testing problem.

Contents · Previous section · Next section

Section 11 of 36

11. Choose neighbourhoods biologically

Set contact radii from measurement resolution and plausible membrane adjacency. Set diffusible neighbourhoods from a justified range or a sensitivity series, not from whichever value gives the clearest map. Preserve barriers such as vessels, cavities or tissue boundaries. Euclidean closeness across an impossible boundary is not biological opportunity.

Contents · Previous section · Next section

Section 12 of 36

12. Protect biological replication

Cells and spots are nested inside sections and specimens. Estimate leading maps within each specimen and show whether locations and directions recur. Thousands of spatial units from one section do not substitute for multiple biological units. A spectacular single-section hotspot is an exploratory observation until replication shows it is not section-specific.

Contents · Previous section · Next section

Section 13 of 36

13. Audit velocity prerequisites

Report spliced and unspliced counting, filtering, gene selection, neighbourhood construction and the velocity model. Inspect whether velocity directions are stable across reasonable settings. Some tissues or protocols provide weak kinetic information. When the velocity fit is poor, the dynamic score should be withheld or clearly labelled as low-confidence.

Contents · Previous section · Next section

Section 14 of 36

14. Freeze current software separately

Cite the 2026 publication for the evaluated method and the exact official repository revision for the implementation used. The repository currently supports common sequencing- and probe-based spatial inputs through cell-by-gene and cell-by-position matrices. That software scope can evolve; a later feature should not be silently attributed to the original study.

Contents · Previous section · Next section

Section 15 of 36

15. Practise with an invented spatial table

This fictional table is a classroom exercise, not CytoSignal output.

Contents · Previous section · Next section

Section 16 of 36

16. Read static and dynamic maps together

A high static score with a near-zero derivative suggests sustained RNA compatibility; a modest score with a positive derivative suggests a possible emerging pattern. These are modelling interpretations, not universal biological states. Display both components and their uncertainty so the derivative does not become an unexplained arrow pasted onto the static heat map.

Contents · Previous section · Next section

Section 17 of 36

17. Inspect ligand and receptor components

For every headline interaction, show ligand expression around the receiver, receptor expression at the receiver and the combined score separately. A high score can be driven by one extreme component. Component maps also reveal whether the proposed sender field is anatomically plausible or merely widespread transcription.

Contents · Previous section · Next section

Section 18 of 36

18. Compare contact and diffusion results

Run plausible alternative classifications or radii for uncertain interactions. A genuine contact-restricted hypothesis should weaken when only distant neighbours contribute, whereas a broad diffusible pattern should not depend on a single adjacent cell. Sensitivity is not a nuisance; it describes the spatial scale on which the claim exists.

Contents · Previous section · Next section

Section 19 of 36

19. Use protein evidence carefully

Proximity ligation, immunostaining or receptor-phosphorylation assays can add molecular evidence. Their spatial resolution, specificity and controls must match the claim. Colocalised proteins strengthen opportunity; receptor-proximal activation strengthens function; perturbation tests necessity. No one assay automatically completes the entire causal chain.

Contents · Previous section · Next section

Section 20 of 36

20. Compare with transparent baselines

Plot ligand expression alone, receptor expression alone, a simple expression product and a distance-weighted product under the same mask. CytoSignal earns interpretive value when its contact rules, gradient structure or velocity extension produce stable information beyond these baselines. Complexity should buy a clearer, more testable hypothesis.

Contents · Previous section · Next section

Section 21 of 36

21. Report uncertainty by layer

Separate uncertainty in segmentation, expression, neighbourhood definition, interaction annotation and velocity. Bootstrap appropriate units, rerun the pipeline and map recurrence of hotspots and derivative signs. A single confidence number cannot tell readers whether instability began in geometry, counts or the kinetic model.

Contents · Previous section · Next section

Section 22 of 36

22. Audit tissue edges

Locations at a section boundary have fewer measured neighbours and may appear unusual. Repeat analyses with an erosion mask or edge covariate, and distinguish physical anatomical boundaries from cut-section artefacts. Never interpret an abrupt signalling edge until the measurement edge has been ruled out.

Contents · Previous section · Next section

Section 23 of 36

23. Audit segmentation and mixed units

Over-segmentation splits transcripts from one cell; under-segmentation merges neighbouring cells. Both alter ligand and receptor pairing. Compare segmentation versions where possible and show conclusions at spot level when cell identity is uncertain. Precision in the prose should never exceed precision in the spatial unit.

Contents · Previous section · Next section

Section 24 of 36

24. Audit abundance and depth

Dense regions provide more potential ligand neighbours and high-depth cells detect more receptors. Plot scores against local cell density, library size and detected genes. Use matched or adjusted analyses to ask whether the spatial pattern remains. A technical opportunity gradient can imitate a biological signalling gradient.

Contents · Previous section · Next section

Section 25 of 36

25. Audit velocity direction

A derivative can change sign when the velocity model, neighbour graph or gene filter changes. Publish sign stability for leading interactions and avoid ranking by magnitude alone. If the static map is stable but the derivative is not, retain the localisation claim and drop the temporal one.

Contents · Previous section · Next section

Section 26 of 36

26. Audit multiple comparisons

Many ligand–receptor pairs across many locations create an enormous search surface. State filters, null construction and correction strategy. Region-level summaries or predeclared interactions can reduce multiplicity, but post-hoc selection after inspecting every map must remain exploratory.

Contents · Previous section · Next section

Section 27 of 36

27. Design a decisive spatial experiment

Choose one replicated hotspot with stable contact or diffusion scale. Confirm ligand and receptor proteins, perturb one component, measure an early receptor-proximal event and map the downstream response inside and outside the predicted range. Include a rescue or competing-ligand control where feasible.

Contents · Previous section · Next section

Section 28 of 36

28. Write the bounded conclusion

Prefer: ‘CytoSignal identified a replicated RNA-based interaction field whose derivative was stable under declared velocity and neighbourhood settings.’ Avoid saying that the tissue was observed sending the signal. The bounded sentence is more useful because it names exactly what the next experiment must test.

Contents · Previous section · Next section

Section 29 of 36

29. Begin with Primary Science maps

Young learners can map where light, heat or water is greatest and ask how the pattern may change. The activity introduces variables, location and prediction without specialised biology. Most importantly, students learn that a map of evidence is not the same thing as a movie of the underlying process.

Contents · Previous section · Next section

Section 30 of 36

30. Use PSLE Science fair tests

Let students vary distance from a source while holding other conditions steady. They can plot an outcome, compare a sharp contact rule with a gradual diffusion rule and identify confounders. This mirrors the reasoning behind choosing spatial neighbourhoods.

Contents · Previous section · Next section

Section 31 of 36

31. Connect Secondary and O-Level Biology

Diffusion, receptors, coordination, gene expression and experimental design provide the biological foundation. CytoSignal adds a data-rich example of how spatial evidence and kinetics can be combined. Students should label what was measured, calculated, predicted and experimentally validated.

Contents · Previous section · Next section

Section 32 of 36

32. Let Computing clarify the derivative

Coordinates, matrices, graphs, convolution, derivatives, version control and reproducible pipelines belong naturally in Computing. A program can calculate a rate of change precisely while the biological interpretation remains uncertain. That distinction is a powerful lesson in responsible STEM.

Contents · Previous section · Next section

Section 33 of 36

33. Guide science tuition productively

Strong science tuition teaches students to link every conclusion to the correct evidence type. A CytoSignal case study can train them to question axes, units, neighbourhoods, controls and alternative explanations instead of memorising a colourful diagram.

Contents · Previous section · Next section

Section 34 of 36

34. Explore STEM and career pathways

Spatial biology connects microscopy, molecular biology, statistics, software engineering and clinical research. Students can explore how multidisciplinary teams build assays, write algorithms, check quality and design validation. Learning the method opens questions; it does not guarantee a particular career outcome.

Contents · Previous section · Next section

Section 35 of 36

35. Choose school opportunities wisely

For school choices, look for current opportunities in practical science, data analysis, coding, research projects and careful communication. Verify programmes on official school pages. A sound science education builds transferable habits rather than familiarity with one specialised package.

Contents · Previous section · Next section

Section 36 of 36

36. Finish with a cheerful question

CytoSignal’s best invitation is wonderfully simple: where should we look next, and what measurement would prove the map wrong? That turns a beautiful spatial pattern into an honest, optimistic experiment.

CytoSignal becomes most useful when its spatial score is treated as a transparent stack of evidence. At the bottom are measured transcripts and coordinates. Above them sit segmentation, ligand–receptor curation, neighbourhood rules and the contact-versus-diffusion classification. The static LRscore combines these choices; VeloCytoSignal then adds splicing information and an RNA-velocity model. A publication should let readers move down that stack whenever a map looks surprising.

Start quality control before calculating a score. Plot library size, detected genes, unspliced fraction, cell area, local density and specimen identity over the tissue. These technical maps often have spatial structure. If a proposed hotspot follows a low-quality edge or a high-depth island, address that relationship directly rather than hiding quality plots in a supplement.

The spatial unit deserves a sentence in every result. Sequencing-based and imaging-based platforms differ in transcript coverage, resolution and segmentation. A cell-by-position matrix may truly describe segmented cells, inferred cell centres or beads that approximate local tissue. The computational interface can accept each format, but the biological nouns in the conclusion must match the measurement.

Contact and diffusion are hypotheses about range. For a contact-dependent route, demonstrate that the leading score relies on immediately adjacent, anatomically compatible units. For a diffusible route, show the score across a planned range of kernels or radii. A route that appears only under one arbitrary bandwidth should be labelled scale-sensitive. This honesty often suggests the most informative imaging experiment.

Velocity adds a tempting sense of motion. Keep it disciplined. RNA velocity estimates change in RNA abundance under kinetic assumptions; it does not track extracellular ligand movement or receptor occupancy. Report the static score, ligand velocity contribution, receptor velocity contribution and combined derivative. When their signs conflict, avoid compressing them into a single story.

Did you know? A positive derivative can coexist with a low present score. That can be biologically interesting because it flags an emerging pattern, but it can also reflect noisy low counts. Require minimum expression, stable velocity direction and cross-specimen recurrence before prioritising such regions. Conversely, a high flat score may represent a sustained programme worthy of validation even without a dramatic derivative.

Spatial null models should preserve what biology makes inevitable. Randomly relocating cells across an entire section destroys anatomy and can create an easy benchmark. Prefer permutations within compatible compartments, rotations or neighbour rewiring that preserve density, and matched-background interactions. State exactly what the null retains and breaks. A null is a scientific claim about what counts as chance.

For multisample comparison, align anatomy carefully without erasing genuine differences. Analyse each specimen first, register only where justified and summarise region-level recurrence. A disease-specific hotspot should not be defined solely by comparing one diseased section with one control section. Specimen-level uncertainty belongs beside every group comparison.

Associated-gene analysis should distinguish local response from cell identity. Include cell-type indicators, region, depth and specimen in diagnostic models. Test whether genes remain associated within a cell type and within anatomical compartments. If an association disappears, it may still be a useful marker of the spatial niche, but it should not be presented as a response to the inferred signal.

Cross-method comparison is especially informative. CellNEST emphasises spatial attention and relay events, COMMOT models collective optimal transport, HoloNet uses multi-view spatial graphs and SpaTalk connects spatial ligand–receptor evidence to target genes. Agreement across different assumptions can strengthen prioritisation. Disagreement maps the part of the conclusion controlled by geometry, prior knowledge or downstream response modelling.

A strong result page can use an evidence card: specimen prevalence; assay and unit; ligand and receptor components; interaction class; neighbourhood range; edge sensitivity; velocity quality; derivative stability; associated genes; protein evidence; perturbation status; and unresolved alternatives. This compact record prevents a heat map from becoming detached from its assumptions.

For validation, match the experiment to the uncertain link. Imaging tests localisation, proximity ligation tests close protein association, receptor phosphorylation tests early activation, reporters test pathway response and perturbation tests necessity. A rescue can strengthen specificity. Spatial controls outside the predicted range are valuable because they test the geometry that motivated the method.

Public communication should preserve the distinction between ‘where RNA evidence is compatible with signalling’ and ‘where signalling occurred’. The first is already exciting: it turns a tissue section into a testable map. Precision does not drain wonder from the result. It tells students and researchers exactly where to aim the next measurement.

Before publication, verify the 2026 paper, PubMed record, official repository and figure repository separately. A repository may retain an earlier preprint citation even after peer-reviewed publication, so use the journal for publication claims and the pinned code revision for implementation. Check all internal links and label the fictional table visibly.

The deepest lesson is that a spatial pattern is an invitation, not a verdict. By asking where a score came from, how its derivative was estimated and which experiment could overturn it, learners practise the joyful discipline that makes modern science trustworthy.

Contents · Previous section · Continue to the Science Learning Hub