Why Science? | iTALK, Ligand–Receptor Categories and Comparative Communication Evidence

Section 1 of 36

1. Start with a candidate conversation map

iTALK matches ligand expression in a nominated source population with receptor expression in a nominated receiver. The resulting edge is a compatibility statement supported by transcript data and a reference pair. It is not a direct recording of a ligand leaving one cell, binding another or changing receiver behaviour.

Contents · Next section

Section 2 of 36

2. Know the publication boundary

The official iTALK repository cites a January 2019 bioRxiv manuscript. BioRxiv labels scientific manuscripts as preprints, so readers should not describe that source as peer reviewed. The repository and preprint can document the proposed workflow; strong biological claims still need independent experimental evidence and current method comparison.

Contents · Previous section · Next section

Section 3 of 36

3. Separate database categories from measurements

iTALK organises known ligand–receptor relations into communication categories. Those categories help readers navigate a large network, but they come from prior knowledge rather than the current specimen. Report the database version and category rules beside expression evidence so an attractive label cannot become proof by itself.

Contents · Previous section · Next section

Section 4 of 36

4. Use highly expressed genes as one route

A highly expressed-gene workflow can highlight abundant candidate signals within a cell group. High abundance may improve detectability, yet it can also favour housekeeping or ubiquitous molecules. Compare prevalence and specificity across groups before calling a pair distinctive.

Contents · Previous section · Next section

Section 5 of 36

5. Use differential expression as another route

A differential-expression workflow can prioritise signals that change between conditions or cell states. The contrast must be defined before ranking, with appropriate biological replicates and covariates. A differential ligand or receptor does not automatically imply differential communication because the other component, cell abundance and physical context may remain unchanged.

Contents · Previous section · Next section

Section 6 of 36

6. Read circle and network plots as indexes

iTALK visualisations compress many rows into chords or edges. Treat them as navigational maps that point back to the underlying table. Width, colour and position are plotting choices; captions should state what each aesthetic encodes and whether unshown edges were filtered or merely absent.

Contents · Previous section · Next section

Section 7 of 36

7. Preserve sender and receiver direction

A ligand-to-receptor edge has an intended direction, even when both populations express both molecules. Store source, ligand, receptor and receiver as explicit columns. A symmetric-looking diagram should not silently turn two directional hypotheses into a proven feedback loop.

Contents · Previous section · Next section

Section 8 of 36

8. Create a specimen-first manifest

Record organism, tissue, condition, donor or animal, batch, assay, preprocessing, cell count, annotation method and comparison group. Thousands of cells from one person improve resolution but do not create thousands of biological replicates. Keep specimen identity available through every summary.

Contents · Previous section · Next section

Section 9 of 36

9. Audit cell annotations before signalling

Cell labels define the network nodes. Check markers, mixed states, doublets, reference mapping and uncertain boundaries before calculating edges. Repeat headline analyses after merging unstable subtypes. A route that disappears under a reasonable label change should be reported as annotation sensitive.

Contents · Previous section · Next section

Section 10 of 36

10. Control ambient RNA and dissociation effects

Ambient transcripts can make a secreted ligand appear outside its true source, while dissociation can induce stress programmes across many cell types. Map contamination and stress indicators to leading edges. A broadly detected cytokine-like transcript may describe preparation rather than tissue signalling.

Contents · Previous section · Next section

Section 11 of 36

11. Choose the contrast before seeing the graph

If every pair of conditions and cell groups is tested until one dramatic chord plot appears, false discovery and narrative bias grow. Pre-specify the primary comparison or label discovery analyses honestly. Hold out specimens or experiments for confirmation.

Contents · Previous section · Next section

Section 12 of 36

12. Define expression and detection rules

State how a gene is considered detected, how group expression is calculated and which minimum fractions are required. Rare signals are especially sensitive to these choices. Show edge persistence over a small prespecified threshold range rather than selecting the threshold that creates the busiest picture.

Contents · Previous section · Next section

Section 13 of 36

13. Freeze identifiers and the interaction resource

Save gene symbols, aliases, orthologue mappings, database rows and package version. An interaction can disappear because an identifier changed or a pair was absent from the resource. That is different from biological absence and deserves a separate missingness code.

Contents · Previous section · Next section

Section 14 of 36

14. Archive a reproducible iTALK run

Retain input matrices, metadata, annotations, contrast design, differential-expression settings, interaction database, filters, long-format results, plot code and software environment. The network image is a view of the analysis, not a reproducible evidence object.

Contents · Previous section · Next section

Section 15 of 36

15. Practise with an invented iTALK audit table

This fictional table is for reasoning practice, not a biological result.

Contents · Previous section · Next section

Section 16 of 36

16. Inspect both molecular components

Open the ligand and receptor distributions for every headline edge. Report detection fraction, abundance, specimen prevalence and uncertainty for both. A composite rank can hide a strong ligand paired with a nearly absent receptor, or a common receptor paired with a source signal present in one donor.

Contents · Previous section · Next section

Section 17 of 36

17. Distinguish abundance from specificity

An abundant pair may be widespread across many sources and receivers; a specific pair may be modest but concentrated. These patterns answer different biological questions. Plot both when possible and avoid using a single colour scale to imply that abundance, uniqueness and confidence are the same thing.

Contents · Previous section · Next section

Section 18 of 36

18. Compare conditions without pseudoreplication

Cells are nested inside biological specimens. Estimate pair support within donors or animals and compare those units using an appropriate design. Cell-level tests alone can produce narrow uncertainty while every apparent difference is driven by one specimen.

Contents · Previous section · Next section

Section 19 of 36

19. Separate cell-state change from cell abundance

A condition can contain more cells of one source type without any cell changing its per-cell expression. Decide whether the question concerns population composition, within-cell state or tissue-level signalling opportunity. Show these components separately instead of calling every network difference rewiring.

Contents · Previous section · Next section

Section 20 of 36

20. Add spatial plausibility

Use imaging or spatial transcriptomics to ask whether source and receiver occupy compatible compartments. Contact-dependent and diffusible mediators need different distance expectations. Proximity supports opportunity, not ligand identity; strong separation can still falsify a proposed contact route.

Contents · Previous section · Next section

Section 21 of 36

21. Use categories without over-generalising

One pair can place an entire edge inside a named category. Inspect which molecules own a category-level pattern and whether results survive removal of a ubiquitous family. Do not state that a whole cytokine or growth-factor programme changed when one database row carries the signal.

Contents · Previous section · Next section

Section 22 of 36

22. Audit interaction-resource sensitivity

Repeat important claims with a high-confidence subset or another justified resource. Mark pairs absent by definition separately from pairs present but unsupported by expression. Agreement strengthens prioritisation; disagreement reveals knowledge uncertainty and should guide manual review.

Contents · Previous section · Next section

Section 23 of 36

23. Audit visual and filtering choices

Vary declared expression cutoffs, plot top-N settings and edge-scaling rules. A diagram can look radically different while the underlying values barely change. Publish a long table and small-multiple sensitivity views so readers can separate analytical change from graphic emphasis.

Contents · Previous section · Next section

Section 24 of 36

24. Compare with transparent baselines

Compare iTALK ranks with a simple expression-product rule and another ligand–receptor method using the same labels and resource where possible. This shows what the differential-expression and visual workflow adds. Agreement across algorithms is sensitivity evidence, not an independent biological replicate.

Contents · Previous section · Next section

Section 25 of 36

25. Validate proteins, contact and response

Confirm ligand and receptor proteins in appropriate compartments, assess whether cells can meet, perturb the proposed route and measure an early receiver event followed by a predicted functional response. Transcript compatibility alone leaves secretion, binding, activation and causality unobserved.

Contents · Previous section · Next section

Section 26 of 36

26. Publish negative and contradictory structure

Include abundant ligands with absent receptors, receptor-rich populations without plausible sources, pairs that fail across specimens and edges removed by reasonable resource changes. These cases calibrate the graph and prevent every coloured chord from being read as success.

Contents · Previous section · Next section

Section 27 of 36

27. Design a falsification experiment

Choose two competing source–receiver explanations and pre-register observations that distinguish them. Include a ligand or receptor blockade, timing and a negative control. A useful network should make different predictions under alternatives, not merely provide a story that can absorb any result.

Contents · Previous section · Next section

Section 28 of 36

28. Write the result at the right strength

Prefer: ‘iTALK prioritised an expression-supported ligand–receptor pair that recurred across specimens and remained stable under declared filters.’ Avoid ‘the cells communicated through this pathway’ until molecular transfer and a receiver response were measured.

Contents · Previous section · Next section

Section 29 of 36

29. Build a classroom model before code

Ask students to build two communication maps from the same fictional expression cards, one based on abundance and one based on condition change, then explain why the highlighted edges differ. Label observations, transformations, assumptions and conclusions separately. The aim is to make every evidence hand-off visible enough for another student to question.

Contents · Previous section · Next section

Section 30 of 36

30. Begin with Primary Science habits

Primary Science already supplies useful foundations for iTALK: careful observation, fair comparison, consistent records and conclusions that fit the evidence. A simple investigation can show why compatible parts create a possibility without proving that an interaction occurred.

Contents · Previous section · Next section

Section 31 of 36

31. Use PSLE Science to practise claim limits

PSLE Science trains students to connect evidence to process without jumping beyond an experiment. With iTALK, ask what was measured, what came from a database, what the algorithm calculated and which new observation could distinguish competing explanations.

Contents · Previous section · Next section

Section 32 of 36

32. Connect Secondary and O-Level Science

Secondary Science and O-Level Biology develop ideas about cells, organisation, transport, experimental design and evaluation. iTALK turns those ideas into a modern data problem. The valuable lesson is disciplined interpretation, not memorising a software menu.

Contents · Previous section · Next section

Section 33 of 36

33. Let Computing support the biology

Computing contributes data structures, scoring, graphs, algorithms, version control and reproducibility. Biology supplies specimens, mechanisms and validation. iTALK shows why correct code is necessary but insufficient: a perfect program can still analyse confounded samples or uncertain labels.

Contents · Previous section · Next section

Section 34 of 36

34. Choose opportunities by verified fit

For school choices and science enrichment, verify current programmes on official school and MOE pages. Look for learning that teaches source checking, visualisation literacy and experimental follow-up together. A fashionable mention of genomics, AI or bioinformatics does not guarantee sustained teaching, admission, mentorship or a particular career outcome.

Contents · Previous section · Next section

Section 35 of 36

35. See connected career families

The reasoning in iTALK appears in single-cell genomics, tumour biology, immunology, bioinformatics, data visualisation and experimental science. Routes may pass through polytechnic, junior college, university or continuing education, with different blends of biology, statistics, computing and communication. Requirements change, so check the responsible institution directly.

Contents · Previous section · Next section

Section 36 of 36

36. Finish with a bounded scientific claim

A defensible conclusion states exactly what iTALK prioritised under which data, resource, settings and specimens. iTALK organises expression-supported ligand–receptor possibilities and visual comparisons; it does not directly measure secretion, binding or response. Report uncertainty, sensitivity, replication, negative results and independent validation together; those boundaries make the conclusion more useful.

The most useful iTALK report keeps its two discovery routes separate. A pair nominated because both genes are abundant answers a different question from a pair nominated because one component changes between conditions. Put the route, contrast and eligibility rule in the edge ledger. Combining the lists without labels can make an ordinary abundant interaction look condition-specific or make a fragile differential result look universally strong.

Communication categories are best treated as indexes into evidence. A category total can rise because many pairs move slightly, because one pair moves strongly, or because a database update added eligible relations. Publish the molecular contribution to every category-level statement and repeat the summary after removing its leading pair. This simple leave-one-family-out check reveals whether the headline belongs to a distributed pattern or a single database row.

Did you know? Two circle plots can show different numbers of edges even when every underlying expression value is identical. Changing the top-N display rule, gene filter or interaction database is enough. That is why the machine-readable edge table is the scientific record and the plot is only a view. Captions should name the filter and identify whether an absent chord was below threshold, outside the database or merely omitted for readability.

Differential-expression evidence inherits the experimental design. A cell-level test may produce a tiny p-value while all apparent change comes from one donor. Preserve specimen identity, use a specimen-aware model and report effect size, direction and donor prevalence. If the dataset cannot support such a model, describe the result as exploratory and test it in independent material rather than allowing cell count to masquerade as replication.

The receiver side deserves the same scrutiny as the source. For every nominated ligand, plot receptor abundance, detection fraction and specimen consistency in the intended receiver. A strong ligand paired with a receptor found in a few uncertain cells is a receptor-limited hypothesis. Keeping both components visible stops an attractive sender story from carrying an unsupported receiver.

Autocrine candidates need single-cell inspection. A broad population can have a high group mean for both ligand and receptor even when different subclusters express the two genes. Show co-detection, subcluster structure and protein localisation before describing self-signalling. When co-expression is absent, the result may instead suggest communication between states hidden inside the original label.

Ambient RNA is a particularly important alternative for abundant secreted ligands. Compare leading source patterns with empty-droplet or contamination estimates where available, inspect whether unrelated cell types show the same transcript and check cellular marker context. A ubiquitous low-level signal may be sample preparation, whereas concentrated expression in a coherent cell state is a stronger starting point.

Condition comparisons should distinguish within-cell expression from composition. A tissue may contain more source cells, a new receiver state or the same populations at different proportions. Report cell abundance, per-cell expression and specimen prevalence separately. These quantities support different biological interpretations, and no single chord-width difference should stand in for all three.

Spatial checks must match mediator biology. Membrane-bound interactions require contact opportunity, secreted proteins may operate across a local field and endocrine signals can travel farther. Use class-specific distance expectations, show coordinate uncertainty and examine whether the nominated cells share a plausible compartment. Proximity is supportive context, not proof of molecule identity or direction.

Negative controls should challenge the actual route to an edge. Shuffle condition labels at the specimen level, test spatially separated cell pairs, inspect high-expression pairs that are absent from the resource and include resource-supported pairs with no receiver response. Each control diagnoses a different failure mode. A single arbitrary cell shuffle cannot calibrate annotation, database and design uncertainty at once.

Method comparison is fairest when inputs are held constant. Use the same counts, cell labels, specimens and interaction resource for iTALK, a simple expression-product baseline and another ligand–receptor workflow. Disagreement then exposes scoring or filtering assumptions. Agreement helps prioritise candidates, but it remains computational agreement because all methods share the same measurements and many of the same priors.

Validation can proceed as a sequence of gates: source ligand protein, receiver receptor or complex, physical opportunity, receptor-proximal activity, downstream response and functional consequence. A failure at one gate should narrow the claim before later assays are interpreted. Rescue or an orthogonal perturbation is especially valuable when several ligands can activate the same receiver programme.

Archive more than the final image. Save the expression object, specimen metadata, annotation hierarchy, contrast model, database snapshot, category mapping, thresholds, full edge table, visualisation settings, software version and random seeds. A future analyst should be able to explain why a particular edge appeared, changed category or vanished after an update.

A useful evidence grade can separate discovery, replicated, spatially plausible, protein-supported and perturbation-supported candidates. Do not collapse these labels into one score unless the weighting is justified and disclosed. Parallel evidence columns let readers see why two similarly ranked pairs may deserve very different follow-up.

An efficient experiment tests alternatives, not just a favourite. Choose two source populations or two ligands that predict different early receiver events, pre-register the sampling times and define results that would favour either route or neither. This turns a communication map into a discriminating scientific instrument rather than a colourful retrospective narrative.

A careful conclusion might say that iTALK prioritised a database-supported ligand–receptor possibility that was reproducible across specimens and stable to declared filters. It should not say the cells communicated through that route until secretion, binding and receiver response are measured. The boundary does not weaken the result; it tells the next experiment exactly what remains unknown.

Contents · Previous section · Continue to the Science Learning Hub