Why Science? | Connectome, Differential Cell-Signalling Topology and Network Evidence

Section 1 of 36

1. Define the word connectome carefully

In this toolkit, a connectome is a graph of potential ligand–receptor signalling between annotated cell groups. It is not the neural connectome of physically traced synapses. State this definition early so readers do not transfer anatomical certainty to an expression-derived network.

Contents · Next section

Section 2 of 36

2. Choose nodes and directed edges

Cell types or states form nodes. A ligand expressed by one group and its receptor expressed by another can create a directed edge under the selected resource and filters. Direction is a modelling convention grounded in molecular roles, not a measurement of molecule travel.

Contents · Previous section · Next section

Section 3 of 36

3. Use flexible ligand–receptor references

Connectome can work with reference sets of known mechanisms. Flexibility helps teams use appropriate resources, but it also makes resource identity part of the result. Archive the exact pair list, evidence scope, complexes, species and modifications.

Contents · Previous section · Next section

Section 4 of 36

4. Calculate topology transparently

The package supports rapid calculation and exploration of signalling networks. Every edge should remain traceable to source expression, target expression, pair identity and filter status. A network graphic is most useful when readers can recover its underlying rows.

Contents · Previous section · Next section

Section 5 of 36

5. Filter for biological readability

Expression and prevalence filters can remove unsupported edges and reduce visual overload. Filters are analytical decisions, not cosmetic settings. Report them in physical or statistical terms and show how many edges were excluded at each step.

Contents · Previous section · Next section

Section 6 of 36

6. Use differential connectomics

Comparative workflows ask how signalling topologies differ across tissues or conditions. A differential edge can arise from expression, abundance, annotation, sampling or resource eligibility. Decompose these contributors before writing that communication changed.

Contents · Previous section · Next section

Section 7 of 36

7. Study hubs without worshipping them

Degree, strength and centrality can highlight prominent senders or receivers. High values may reflect ubiquitous molecules or large cell populations. Test whether a hub persists after removing high-degree ligand families and balancing sampling.

Contents · Previous section · Next section

Section 8 of 36

8. Create a matched data manifest

Record tissue, condition, specimen, assay, preprocessing, annotation and cell counts. Comparative topology requires matched definitions. If one dataset lacks a cell type, network density and normalisation change even when shared biology is identical.

Contents · Previous section · Next section

Section 9 of 36

9. Audit cell labels at several resolutions

Fine subtypes reveal specialised edges but increase sparsity; coarse labels stabilise expression but mix states. Compare a planned hierarchy and report which conclusions survive. Avoid presenting a resolution-specific edge as universal.

Contents · Previous section · Next section

Section 10 of 36

10. Inspect ligand and receptor distributions

Group averages can hide zero inflation or a rare high-expressing subpopulation. Plot expressing-cell fractions, specimen distributions and relevant subunits. An edge supported by one small cluster of cells should be described at that scale.

Contents · Previous section · Next section

Section 11 of 36

11. Separate cells from biological replicates

Pooled single-cell estimates are descriptive of the sampled cells. Condition claims require donor or animal replication. Calculate per-specimen support or use a model that respects nested cells, then show heterogeneity rather than only a pooled graph.

Contents · Previous section · Next section

Section 12 of 36

12. Respect complexes and families

If a receptor requires multiple subunits, specify how completeness is handled. If a resource lists families, show the actual genes that support an edge. A family label should not turn partial transcript evidence into a functional complex.

Contents · Previous section · Next section

Section 13 of 36

13. Preserve missing cell populations

Absence of a node may reflect biology, sampling or annotation. Do not quietly remove missing populations from denominators in one condition. Use a declared shared universe and mark structurally missing comparisons.

Contents · Previous section · Next section

Section 14 of 36

14. Archive the full edge ledger

Save input matrices, metadata, cell hierarchy, specimen IDs, ligand–receptor resource, thresholds, package version, network tables, comparison rules and plotting settings. Interactive exploration helps discovery but is not a substitute for a versioned ledger.

Contents · Previous section · Next section

Section 15 of 36

15. Practise with an invented topology audit table

This fictional table shows why a differential edge needs several checks.

Contents · Previous section · Next section

Section 16 of 36

16. Read an edge before the whole graph

Inspect ligand, receptor, cell pair, specimen prevalence and threshold distance from cutoff. Network position is secondary to molecular support. A beautifully central edge with weak component evidence is still weak.

Contents · Previous section · Next section

Section 17 of 36

17. Compare topology on matched edge sets

Different resources or missing genes can change the set of possible edges. For condition comparisons, distinguish changes among shared eligible edges from edges available in only one analysis. Otherwise set membership can masquerade as biology.

Contents · Previous section · Next section

Section 18 of 36

18. Normalise network summaries cautiously

Total edge count, summed weight and centrality depend on graph density. Explain any scaling and show raw values. A denser graph is not automatically a more communicative tissue; it may simply use a broader resource or lower threshold.

Contents · Previous section · Next section

Section 19 of 36

19. Interpret differential edges as decompositions

For every changed edge, ask whether the ligand moved, the receptor moved, detection changed, the cell types changed in abundance or the annotation boundary shifted. The phrase ‘differential communication’ compresses these distinct explanations.

Contents · Previous section · Next section

Section 20 of 36

20. Use visualisation as an audit tool

Circles, matrices and chord diagrams reveal patterns, but also hide overlap and uncertainty. Pair overview graphics with sortable edge tables, specimen panels and clear legends. Avoid edge-width scales that make small numerical differences look dramatic.

Contents · Previous section · Next section

Section 21 of 36

21. Add spatial and protein evidence

Use spatial proximity appropriate to the signal range, then verify ligand and receptor proteins. These checks move the graph from transcript compatibility toward biological opportunity. Binding, activation and response still require additional evidence.

Contents · Previous section · Next section

Section 22 of 36

22. Audit resource sensitivity

Rebuild the network with a high-confidence subset or alternative reference. Track edges lost because the resource omitted them versus edges retained but re-ranked. Stable topology across resources strengthens prioritisation.

Contents · Previous section · Next section

Section 23 of 36

23. Audit threshold sensitivity

Vary detection and expression thresholds over a small prespecified range. Summarise edge persistence and rank intervals. A hub that exists only at one threshold should not anchor the headline.

Contents · Previous section · Next section

Section 24 of 36

24. Compare with another method

Run a second transparent communication method with identical inputs. Agreement can identify a stable shortlist; disagreement reveals scoring assumptions. Neither substitutes for new biological measurements because both reuse the same transcript data.

Contents · Previous section · Next section

Section 25 of 36

25. Validate comparative claims

Use matched specimens, blinded image quantification, protein assays and perturbations. For a differential edge, test both conditions in the same experimental system and measure a predicted downstream response, not merely another expression snapshot.

Contents · Previous section · Next section

Section 26 of 36

26. Publish conserved and absent structure

Report conserved edges, reproducible nulls and populations that never form supported connections. These results calibrate differential claims and may reveal stable tissue infrastructure that flashy contrasts overlook.

Contents · Previous section · Next section

Section 27 of 36

27. Turn topology into a decision

Select one robust differential edge, one composition-sensitive edge and one resource-sensitive edge for matched validation. The trio tests whether the network ranking is calibrated and where assumptions dominate.

Contents · Previous section · Next section

Section 28 of 36

28. Use careful connectome language

Write that Connectome ‘calculated an expression-supported signalling topology and identified a replicated differential edge.’ Do not call it physical wiring or direct communication unless experiments measured those features.

Contents · Previous section · Next section

Section 29 of 36

29. Build a classroom model before code

Let students build two graphs from the same cell cards using different thresholds, then compare which apparent hubs are stable and which were created by the rule. Label observations, transformations, assumptions and conclusions separately. The aim is not to imitate specialist software; it is to make each evidence hand-off visible enough for another student to challenge.

Contents · Previous section · Next section

Section 30 of 36

30. Begin with Primary Science habits

Primary Science already supplies the foundations for Connectome: 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 the experiment. With Connectome, ask what was measured, what came from a knowledge resource, what the algorithm calculated and which new observation could separate 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. Connectome turns these 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, graphs, algorithms, ranking, version control and reproducibility. Biology supplies specimens, mechanisms and validation. Connectome shows why correct code is necessary but insufficient: a perfect program can still analyse confounded samples or doubtful 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 teaching that makes data provenance and comparative reasoning visible, not only attractive network pictures. 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 Connectome appears in single-cell genomics, network biology, tissue engineering, bioinformatics, data visualisation, statistics 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 Connectome prioritised under which data, resource, settings and specimens. Connectome calculates ligand–receptor signalling topologies from expression and reference interactions; it does not directly map physical contact, molecular transfer or causal response. Report uncertainty, sensitivity, replication, negative results and independent validation together; those boundaries make the conclusion more useful.

Connectome makes cell-signalling topology inspectable, but the word “connectome” can overpromise. In neuroscience it may evoke physically traced wiring. Here it denotes a computational graph of ligand–receptor possibilities supported by group-level expression and a reference set. Put that definition in captions and methods so network language does not silently acquire anatomical certainty.

Every edge should carry provenance. Record source cell, ligand, receptor, target cell, expression values, detection fractions, complex rule, resource record, threshold status and specimen prevalence. If a differential edge cannot be traced to these fields, the network is too compressed for scientific interpretation. Provide machine-readable long tables with figures.

Flexible resources are a strength because different tissues and species need different knowledge bases. They are also a major source of variation. Run a high-confidence subset and one justified alternative where possible. Mark interactions absent from a resource separately from interactions present but unsupported by expression. Those are different kinds of missingness.

Did you know? A network can gain a new hub without any cell changing expression if a broader ligand–receptor resource is substituted. The graph’s eligible edge set expanded. This is why resource version belongs beside biological condition in every comparison and why topology should never be compared across undocumented databases.

Filtering is part of the scientific model. Minimum expression and prevalence decide who can send or receive. Show edge counts before and after each filter and repeat important claims across a small prespecified range. A centrality result that collapses after a minor threshold change belongs in a sensitivity tier.

Group averages need distributional context. A moderately expressed receptor in every receiver can mean something different from the same mean generated by a tiny high-expression subset. Display fractions, quantiles and specimen panels. If biology points to a rare responder state, model that state rather than letting the average blur it.

Comparative connectomics must preserve biological units. Build per-specimen networks or estimate edge support with nested models, then compare conditions across donors or animals. Thousands of cells improve estimation within a specimen but do not multiply the number of independent cases. Report prevalence and heterogeneity alongside pooled effect summaries.

Cell composition can change network topology even when per-cell expression is stable. More abundant source cells may produce more total ligand opportunities, while the group mean remains unchanged. Decide whether the question concerns per-cell state, population abundance or tissue-level opportunity, and choose weights accordingly. Do not switch interpretations midstream.

Node identity should be harmonised across datasets with an explicit mapping. A disease-specific state may lack a healthy counterpart, and forcing it into the nearest label can manufacture a differential edge. Use hierarchical comparisons: shared broad types first, supported subtypes second, and condition-specific nodes as structurally unmatched.

Centrality is relative to graph construction. Degree, strength, betweenness and other measures answer different topological questions. Name the metric and show whether results survive removal of ubiquitous ligands or high-degree receptor families. Avoid calling a cell type a universal hub when its prominence belongs to one resource or threshold.

Differential edges should be decomposed into ligand change, receptor change, detection change, cell-number change, label change and resource eligibility. A waterfall or audit table can make these contributors visible. The phrase “communication increased” is too strong when only one transcript component rose or one cell population became more abundant.

Visual exploration is excellent for discovery. It is poor as the sole record because dense diagrams hide overlapping edges and scale choices. Pair network views with sortable tables, small-multiple specimen plots and a clear explanation of width, colour and filtering. The figure should invite inspection rather than settle the argument by appearance.

Use spatial evidence appropriate to mediator biology. Membrane-bound interactions need close contact; secreted factors may cross several cell diameters; systemic signals can operate over longer distances. Apply per-class expectations and report coordinate uncertainty. A single nearest-neighbour rule cannot validate every edge in the same graph.

Compare Connectome with a simple expression-product baseline and another method using the same resource. This reveals whether its topology and differential workflow add stable information. Agreement among algorithms can prioritise candidates, but shared counts, labels and databases mean it is not independent biological confirmation.

Negative controls should include shuffled condition labels at the specimen level, spatially incompatible cell pairs, high-expression but non-interacting gene pairs and resource subsets with known exclusions. Controls must preserve the dependence structure relevant to the test. Cell-level shuffling can create false precision when specimens are the true units.

Validation should test both topology and mechanism. Confirm proteins and location, perturb the candidate edge, measure an early receiver event and observe the predicted functional outcome. For differential claims, apply the same assay in both conditions with blinded quantification. A network difference that lacks a receiver consequence may still be descriptive but should not be called functional rewiring.

Archive the input object, metadata dictionary, label hierarchy, resource, filters, per-specimen edges, aggregate graph, comparison formulas, centrality metrics, sensitivity runs, plots and software environment. Reproducibility means a future analyst can regenerate the edge ledger and explain every disappearance, not merely redraw a similar-looking chord diagram.

A careful conclusion might read: “Connectome identified a differential expression-supported edge from population S to T that recurred in five of six condition-A specimens, one of six controls and remained stable across declared resource, threshold and annotation checks.” Reserve ‘rewired signalling’ for stronger molecular and functional evidence.

Finish with a portfolio of tests: one robust edge, one threshold-sensitive edge and one resource-sensitive edge. Validate all three under the same assay. Their successes and failures reveal whether the network is calibrated and which assumptions matter most. That learning is more valuable than selecting only the easiest candidate to confirm.

Contents · Previous section · Continue to the Science Learning Hub