Why Science? | ICELLNET, Multisubunit Ligand–Receptor Networks and Immune-Crosstalk Evidence

Section 1 of 36

1. Start with a human communication network

ICELLNET links a cell population of interest to other human cell populations through curated ligand–receptor relations and expression-derived communication scores. The network is a structured inventory of plausible routes. It is not a movie of molecules moving between cells.

Contents · Next section

Section 2 of 36

2. Use manual curation as evidence, not certainty

The database is biologically curated from human studies, which improves interpretability. Curation still reflects published coverage, experimental contexts and selection rules. Record the interaction record and source evidence for every headline pair.

Contents · Previous section · Next section

Section 3 of 36

3. Respect multiple ligand and receptor subunits

ICELLNET explicitly considers complexes with several subunits. A route requiring two receptor chains should not be treated as complete when only one is detected. State the completeness rule, missing subunits and how low expression is handled.

Contents · Previous section · Next section

Section 4 of 36

4. Know the v1 and v2 resource difference

The 2024 ICELLNET v2 report describes an expansion from 380 to 1,669 curated interactions and additional communication families. A larger eligible resource can change edge counts and hubs without any biology changing. Freeze the exact release and do not compare undocumented versions.

Contents · Previous section · Next section

Section 5 of 36

5. Group interactions into interpretable families

Families such as immune crosstalk, adhesion or Wnt-related communication help readers navigate many pairs. Family scores are aggregates. Return to the individual interactions and subunits before writing that an entire programme is activated.

Contents · Previous section · Next section

Section 6 of 36

6. Quantify communication scores cautiously

A communication score combines expression information under ICELLNET’s rules. It organises candidates but is not a biochemical rate. Show the source and receiver components, scale and denominator so a large number is not mistaken for signal flux.

Contents · Previous section · Next section

Section 7 of 36

7. Use reference profiles with explicit boundaries

ICELLNET can connect a population of interest with reference human cell profiles. Reference comparisons can reveal plausible partners, but they may differ in tissue, activation state, assay and laboratory. Validate important receivers in matched material.

Contents · Previous section · Next section

Section 8 of 36

8. Create a human-specimen manifest

Record tissue, condition, donor, batch, assay, sample processing, cell counts, sorted or single-cell input, annotation method and reference profiles. A network built from one donor and many reference cells remains a one-donor observation.

Contents · Previous section · Next section

Section 9 of 36

9. Audit the cell population of interest

Confirm identity, purity, activation state and contamination of the focal population. If it is sorted, document markers and gating; if it is single-cell derived, examine doublets and mixed clusters. A mislabelled emitter changes every outgoing score.

Contents · Previous section · Next section

Section 10 of 36

10. Match reference and sample states

A resting reference cell and an inflamed tissue cell may share a broad name but not a transcriptomic state. Use matched references when possible and label unmatched comparisons as exploratory. Do not interpret reference distance as disease communication.

Contents · Previous section · Next section

Section 11 of 36

11. Control immune activation confounders

Interferon, stress and cytokine programmes can rise across many cells after processing or inflammation. Inspect whether one generic programme drives multiple families. Use specimen replication and orthogonal protein measurements to separate biology from handling.

Contents · Previous section · Next section

Section 12 of 36

12. Define complex completeness and detection

State minimum expression and prevalence for every required subunit. Report complete, partial and missing complexes separately. An aggregate score should not hide a receptor chain detected in only one cell or one specimen.

Contents · Previous section · Next section

Section 13 of 36

13. Freeze human gene identifiers and database rows

Archive symbols, aliases, retired names, interaction IDs, subunit composition, evidence source and release. Database absence is not biological impossibility; sample absence is not the same as a failed identifier map.

Contents · Previous section · Next section

Section 14 of 36

14. Archive a reproducible ICELLNET run

Save expression inputs, metadata, cell labels, reference profiles, database release, thresholds, complex rules, scoring settings, long-format results, visualisation code and software environment. Figures should be regenerable from the edge ledger.

Contents · Previous section · Next section

Section 15 of 36

15. Practise with an invented complex audit table

This fictional table is for reasoning practice, not ICELLNET output.

Contents · Previous section · Next section

Section 16 of 36

16. Read communication scores pair by pair

Display ligand, every receptor subunit, detection fractions and specimen prevalence for top scores. When one component owns the aggregate, name that dependence. Keep the communication score and the molecular components in adjacent columns.

Contents · Previous section · Next section

Section 17 of 36

17. Inspect outgoing and incoming summaries

Outgoing scores can highlight candidate emitters; incoming scores can highlight candidate receivers. These are relative to the included populations and resource. Add or remove a cell type and the network context changes, so use matched universes for comparisons.

Contents · Previous section · Next section

Section 18 of 36

18. Interpret family-level patterns carefully

A strong family can arise from many modest pairs or one dominant interaction. Publish contribution tables and sensitivity to that pair’s removal. The family label should organise evidence, not replace it.

Contents · Previous section · Next section

Section 19 of 36

19. Compare conditions at the specimen level

Build or score networks for independent donors or animals, then compare conditions. Separate expression change, cell abundance and reference mismatch. A pooled difference should not be described as replicated immune rewiring.

Contents · Previous section · Next section

Section 20 of 36

20. Distinguish contact, secreted and matrix routes

Different interaction families imply different spatial constraints. Adhesion and membrane-bound pairs require close contact; secreted molecules may act over distance; extracellular-matrix routes depend on localisation. Apply mechanism-specific plausibility checks.

Contents · Previous section · Next section

Section 21 of 36

21. Add protein and spatial evidence

Use flow cytometry, imaging, proteomics or other appropriate assays to confirm complete complexes and localisation. Spatial co-occurrence supports opportunity, while missing required protein can falsify a transcript-supported edge.

Contents · Previous section · Next section

Section 22 of 36

22. Audit database-release sensitivity

Compare the chosen release with a stable high-confidence subset or another mapped resource. A v2-only edge may be valuable, but its novelty comes from resource expansion and must not be presented as a new biological change in old samples.

Contents · Previous section · Next section

Section 23 of 36

23. Audit reference-cell sensitivity

Repeat the analysis with matched tissue cells, alternate references or no external reference where possible. Track edges that persist. Reference sensitivity is a property of the inference, not a reason to hide an inconvenient result.

Contents · Previous section · Next section

Section 24 of 36

24. Compare transparent baselines

Use a simple complete-complex expression rule and another ligand–receptor method with the same eligible pairs. ICELLNET adds value when curation, complex handling and reference comparisons yield stable candidates that predict new observations.

Contents · Previous section · Next section

Section 25 of 36

25. Validate immune-crosstalk mechanisms

Confirm ligand production, full receptor complex, proximity, proximal signalling and downstream function. Block the proposed route and include an alternative ligand or receptor control. Immune networks are redundant; one perturbation may expose compensation rather than disprove all communication.

Contents · Previous section · Next section

Section 26 of 36

26. Publish partial complexes and null families

Show interactions excluded by missing subunits, families not supported, and prominent scores that fail across specimens. These negatives demonstrate that complex rules were applied consistently and help calibrate the positive network.

Contents · Previous section · Next section

Section 27 of 36

27. Design a complex-specific falsification test

Choose a leading interaction with an essential auxiliary subunit. Test whether loss or blockade of that subunit removes the proximal receiver response. This distinguishes complete-complex biology from a graph built on one attractive component.

Contents · Previous section · Next section

Section 28 of 36

28. Write a bounded ICELLNET result

Prefer: ‘ICELLNET prioritised a curated, complete ligand–receptor complex whose score and specimen support were stable across declared database and reference checks.’ Avoid claiming immune crosstalk was activated until protein, timing and response evidence support it.

Contents · Previous section · Next section

Section 29 of 36

29. Build a classroom model before code

Use cards for ligand and receptor subunits, then let students decide whether a complex is complete before any network edge may be drawn. 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 ICELLNET: 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 ICELLNET, 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. ICELLNET 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. ICELLNET 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 treats molecular complexes, human evidence curation and validation as equally important. 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 ICELLNET appears in immunology, cancer biology, molecular pathology, bioinformatics, drug discovery and systems biology. 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 ICELLNET prioritised under which data, resource, settings and specimens. ICELLNET calculates expression-supported communication scores from curated human interactions and complex rules; it does not directly measure complex formation or signal transmission. Report uncertainty, sensitivity, replication, negative results and independent validation together; those boundaries make the conclusion more useful.

ICELLNET’s curated human resource is most valuable when curation and specimen evidence remain separate. For each headline interaction, record the database release, interaction ID, evidence scope, all subunits and the current sample’s expression support. Manual curation raises interpretability, but a literature-supported relation may have been observed in another tissue, activation state or experimental system.

Multi-subunit complexes require an AND-rule unless the biological record says otherwise. Report every required chain, detection fraction and specimen prevalence. A strong alpha chain cannot compensate silently for a missing beta chain. If the software’s default aggregation differs from the biological question, document the change and test both specifications.

Did you know? Expanding a ligand–receptor database can make the same old sample look more connected. ICELLNET v2 reported many more curated interactions than the first publication. A new edge in a reanalysis may therefore reflect a newly eligible hypothesis, not new biology. Resource version belongs beside condition and specimen in every network comparison.

Family-level summaries are useful for orientation but can conceal ownership. Decompose immune-crosstalk, adhesion, Wnt-related or other families into contributing pairs and repeat the result after removing the largest contributor. A distributed family signal and a single dominant pair demand different validation plans and different language.

Reference profiles create a practical comparison universe, yet they may differ from the focal sample in tissue, laboratory, assay and activation. Quantify similarity before treating a reference as a plausible receiver, and repeat important results with a matched or alternate reference when possible. Unmatched reference results are hypothesis generators, not reconstructed tissue conversations.

The population of interest must be audited carefully. For sorted cells, preserve gating, purity and viability information; for single-cell clusters, inspect markers, doublets and mixed states. Contamination in the focal emitter can create many apparently coherent outgoing edges because the same incorrect identity is reused throughout the network.

Generic inflammation can dominate human communication maps. Plot interferon, stress and immediate-early programmes by specimen and ask whether one shared state drives several families. A network that disappears after accounting for processing stress should be framed as preparation-sensitive, while a reproducible tissue-specific pattern deserves stronger follow-up.

Outgoing and incoming scores are relative to included cell populations. Adding a new reference type changes the comparison context even if expression in the original groups is untouched. Use a matched population universe for condition contrasts and declare any structurally missing group. Relative prominence should never be narrated as an absolute secretion rate.

Communication scores are computational aggregates, not kinetic constants. State the expression scale, aggregation rule, complex handling and any normalisation across cell types. Readers should be able to reproduce why a pair obtained its value. Units without such documentation can look quantitative while remaining biologically ambiguous.

Specimen replication must precede condition claims. Compute or summarise candidate support within donors and show prevalence and heterogeneity. A pooled edge driven by one inflamed specimen is different from a moderate edge present in every case. Cell count cannot replace the number of independent people or animals.

Spatial constraints vary by interaction class. Adhesion and membrane-bound pairs need close contact, extracellular-matrix relations depend on localisation and secreted mediators may act across a field. Apply family-appropriate plausibility rules and show actual distance distributions when coordinates are available. One universal neighbourhood radius is rarely defensible.

Database sensitivity should include a stable high-confidence subset and, where feasible, another mapped resource. Track interactions lost because of missing database membership separately from interactions retained but unsupported by sample expression. This distinction turns disagreement into information about knowledge coverage instead of a hidden nuisance.

Use a plain complete-complex expression baseline. ICELLNET adds value if curation, family organisation and reference comparison produce candidates that remain stable and predict independent evidence. Comparing methods with different cell labels or resources does not isolate method behaviour, so hold inputs constant whenever possible.

Negative results are especially informative for complex-aware analysis. Publish partial complexes, families with no eligible complete edge and prominent scores that fail across donors or protein assays. These examples demonstrate that completeness rules were applied consistently and help readers interpret positive candidates.

Validation should confirm ligand production, every essential receptor subunit, appropriate localisation, a proximal pathway event and a receiver response. Blocking an auxiliary chain can be a sharp complex-specific test. In redundant immune systems, loss of one route may trigger compensation, so measure early molecular events before interpreting later phenotype.

Archive focal-cell inputs, reference profiles, label definitions, database release, subunit logic, thresholds, long-format scores, family contributions, sensitivity runs and software environment. Keep the v1-versus-v2 distinction explicit in filenames and methods. A future reanalysis should be able to identify whether a changed conclusion came from biology, reference choice or resource expansion.

A defensible statement says ICELLNET prioritised a curated, complete interaction whose score recurred across specimens and survived declared database and reference checks. It should not say immune crosstalk was activated until timed protein and response evidence supports that claim. Clear limits make the candidate network more actionable.

Contents · Previous section · Continue to the Science Learning Hub