Why Science? | scConnect, Multi-Directional Graphs and Molecular-Ligand Evidence

Section 1 of 36

1. Start with a putative interaction graph

scConnect estimates ligand expression in one cell population and receptor expression in another, then places compatible pairs into a directed graph. The edge is putative: it combines transcript measurements with prior pharmacology knowledge. It is not a direct recording of molecule release, diffusion, binding or receiver response.

Contents · Next section

Section 2 of 36

2. Know what multi-directional means

The graph can preserve different directions and molecular routes between populations. Direction follows the nominated ligand-bearing source and receptor-bearing receiver; it does not prove causal flow. Store source, ligand, receptor and receiver explicitly so reciprocal candidates are not collapsed into one undirected relationship.

Contents · Previous section · Next section

Section 3 of 36

3. Treat the Guide to Pharmacology as an input

The published workflow draws interaction knowledge from the IUPHAR/BPS Guide to Pharmacology. That resource defines eligible molecular partners before the specimen is analysed. Record its version, species mapping and relevant interaction records, and distinguish database absence from lack of expression in the sample.

Contents · Previous section · Next section

Section 4 of 36

4. Include molecular ligands with care

A distinctive scConnect feature estimates some molecular ligands through genes involved in synthesis, transport and reuptake. This can represent neurotransmitters and other non-protein signals more faithfully than one transcript alone. It also introduces an explicit biochemical model whose components and weights need inspection.

Contents · Previous section · Next section

Section 5 of 36

5. Separate production machinery from ligand quantity

Expression of enzymes and transporters can support capacity to produce or handle a molecular ligand, but it does not directly measure concentration, release or turnover. Metabolomics, secretion assays and receptor-proximal readouts are needed when the hypothesis depends on actual molecular availability.

Contents · Previous section · Next section

Section 6 of 36

6. Use specificity as a comparative quantity

scConnect uses label permutations to estimate how specifically ligands and receptors associate with cell types. Specificity depends on the included populations, sample composition and permutation scheme. Report the comparison universe and null model rather than presenting a specificity score as an intrinsic property of the molecule.

Contents · Previous section · Next section

Section 7 of 36

7. Keep exploratory purpose visible

The primary paper describes scConnect as an exploratory method for contextual analysis and hypothesis generation. That is a strength: it organises a complex system into testable routes. It is not a licence to replace protein, spatial or perturbation evidence with graph structure.

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 and annotation method. Preserve specimen identity through aggregation. Many cells from one specimen improve resolution but do not create independent biological replication.

Contents · Previous section · Next section

Section 9 of 36

9. Audit labels before graph construction

Cell-type labels define the nodes and the groups used in permutations. Check marker coherence, doublets, transitional states and uncertain reference mappings. Repeat leading results at a coarser annotation level so a fragile subtype boundary cannot silently own the graph.

Contents · Previous section · Next section

Section 10 of 36

10. Inspect cell counts and group balance

Very small populations have unstable averages and permutation behaviour, while large groups can dominate summary statistics. Report cells per population and specimen. Use balanced resampling as a sensitivity analysis where appropriate, but keep the original biological composition visible.

Contents · Previous section · Next section

Section 11 of 36

11. Control ambient RNA and stress

Ambient transcripts may distribute a secreted ligand across unrelated droplets, and dissociation can create broad stress programmes. Map contamination and stress indicators to leading source populations. For molecular ligands, also inspect whether biosynthetic genes form a coherent cell programme rather than isolated noisy detections.

Contents · Previous section · Next section

Section 12 of 36

12. Freeze preprocessing choices

Normalisation, feature filtering and minimum expression rules affect group estimates. Keep raw counts, document transformations and run one planned alternative. Do not tune preprocessing until a desired route appears; record why each choice fits the assay and question.

Contents · Previous section · Next section

Section 13 of 36

13. Archive identifiers and species mappings

Save gene symbols, aliases, orthologues, ligand definitions, receptor records and any user additions. A route can disappear because an identifier changed rather than because biology was absent. Give each missing state a traceable reason.

Contents · Previous section · Next section

Section 14 of 36

14. Archive a reproducible scConnect run

Retain the expression object, metadata, annotations, database build, molecular-ligand gene sets, permutation settings, graph tables, plotting code, software version and seeds. The interactive graph is a viewing layer, not the full scientific record.

Contents · Previous section · Next section

Section 15 of 36

15. Practise with an invented graph audit table

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

Contents · Previous section · Next section

Section 16 of 36

16. Inspect ligand and receptor distributions

Open the cell-level distributions behind every leading edge. Report abundance, detection fraction, specimen prevalence and uncertainty for both components. A graph weight can hide a strong ligand paired with a receptor detected in very few cells.

Contents · Previous section · Next section

Section 17 of 36

17. Inspect every molecular-ligand component

For an ensemble-derived ligand, display synthesis, transport, vesicle, degradation or reuptake genes according to the declared model. Identify the component that limits the estimate. A complete biochemical route is more persuasive than one abundant enzyme standing in for the whole process.

Contents · Previous section · Next section

Section 18 of 36

18. Read graph degree cautiously

High degree may reflect biological breadth, ubiquitous expression or a database with many annotated partners. Compare degree with expression, specificity and resource coverage. Remove high-frequency molecules in a sensitivity run before naming a cell population a communication hub.

Contents · Previous section · Next section

Section 19 of 36

19. Preserve biological replicates

Estimate edge support within each donor or animal and summarise recurrence across specimens. A pooled graph can look stable while being driven by one sample. Use the specimen, not the cell, as the independent unit for population-level contrasts.

Contents · Previous section · Next section

Section 20 of 36

20. Separate abundance from cell composition

A condition may contain more source cells without increasing expression per cell. Report cell abundance, per-cell ligand machinery and receiver support separately. These components answer different questions and should not be collapsed into the phrase ‘more communication’.

Contents · Previous section · Next section

Section 21 of 36

21. Add spatial and circuit plausibility

Ask whether nominated source and receiver populations occupy compatible tissue compartments or neural circuits. Contact-dependent routes need close opportunity; diffusible and synaptic signals require different checks. Spatial agreement supports feasibility but does not identify the transmitted molecule by itself.

Contents · Previous section · Next section

Section 22 of 36

22. Audit the permutation null

State what labels are shuffled, which structure is preserved and how many permutations are used. A null that mixes specimens or destroys condition structure may be biologically irrelevant. Compare results across one justified alternative null for headline specificity claims.

Contents · Previous section · Next section

Section 23 of 36

23. Audit pharmacology-resource sensitivity

Repeat important protein routes with a high-confidence subset or another curated interaction resource mapped to the same genes. For molecular ligands, verify the biochemical definition against current pharmacology knowledge. Mark resource-sensitive edges clearly.

Contents · Previous section · Next section

Section 24 of 36

24. Compare with transparent baselines

Use a simple ligand–receptor expression product and another method with identical labels and eligible pairs. For molecular ligands, compare the ensemble estimate with direct metabolite or protein data when available. Complexity earns trust when it improves discriminating predictions.

Contents · Previous section · Next section

Section 25 of 36

25. Validate production, receptor and response

Measure the ligand or its production machinery, confirm the receptor protein, test physical opportunity, perturb the source or receptor and record an early receiver event followed by a later phenotype. Each step closes a different inferential gap.

Contents · Previous section · Next section

Section 26 of 36

26. Publish contradictions and nulls

Show high-specificity routes that lack receptor protein, molecular-ligand models that fail metabolite measurement, edges unstable across specimens and known interactions not recovered. These examples calibrate interpretation better than a graph containing only attractive positives.

Contents · Previous section · Next section

Section 27 of 36

27. Design an alternative-sensitive experiment

Choose two routes that predict different early receiver events or respond differently to synthesis versus receptor blockade. Pre-register measurements, timing and a result that rejects both. A useful graph should help discriminate mechanisms.

Contents · Previous section · Next section

Section 28 of 36

28. Write the result at the right strength

Prefer: ‘scConnect prioritised a resource-supported directed route with reproducible ligand, receptor and specificity evidence.’ Avoid ‘the cells signalled through this route’ until production, transfer and receiver activation are measured.

Contents · Previous section · Next section

Section 29 of 36

29. Build a classroom model before code

Give students expression cards for a protein ligand and an enzyme-built molecular ligand, then ask why the two require different evidence chains. 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 scConnect: 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 scConnect, 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. scConnect 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. scConnect 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 programmes that connect molecular biology, graph reasoning, pharmacology resources and laboratory validation. 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 scConnect appears in neuroscience, pharmacology, single-cell genomics, bioinformatics, network analysis and experimental 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 scConnect prioritised under which data, resource, settings and specimens. scConnect builds transcript- and resource-supported interaction graphs; it does not directly observe ligand production, release, binding or causation. Report uncertainty, sensitivity, replication, negative results and independent validation together; those boundaries make the conclusion more useful.

The most important scConnect distinction is between a protein ligand represented by its encoding transcript and a molecular ligand represented through a set of genes. Publish that class beside every edge. The latter may integrate synthesis, packaging, transport or reuptake machinery, so its score is a model of biochemical capacity rather than an assay of the ligand itself.

Did you know? A neuron can express the receptor for a neurotransmitter while the source cell expresses several relevant enzymes, yet no release may occur during the sampled state. Vesicle loading, stimulation, degradation and local concentration still matter. This is why molecular-ligand candidates benefit from metabolomics, secretion or electrophysiological evidence rather than transcript confirmation alone.

Specificity is conditional on the populations in the analysis. Remove one broadly expressing cell type and the remaining source may appear more specific without changing a single count. Keep a shared population universe for comparisons and show structurally missing groups. A permutation score should never be detached from the labels and specimens that defined its null.

The multi-directional graph should retain edge provenance. Include source population, ligand definition, receptor, receiver, component genes, group expression, specificity, database record, specimen prevalence and validation status. A reader should be able to move from an attractive network line back to the measurements and assumptions that created it.

Graph density is sensitive to database breadth and expression filters. Show edge counts before and after each rule, then recalculate centrality across a planned threshold range. A hub that exists only in the densest graph is a construction-sensitive lead, not a stable biological controller.

Molecular-ligand gene sets can share enzymes or transporters. One abundant component may support several ligand estimates, producing correlated edges. Publish component-overlap matrices and rerun summaries after removing common machinery. This reveals whether the network contains distinct biochemical routes or repeated consequences of one generic gene programme.

Use specimen-aware permutations when conditions or donors matter. Shuffling labels across all cells can erase real donor structure and create a misleading null. Preserve the nesting and pairing relevant to the question, and explain why the chosen randomisation represents absence of the specific effect being tested.

Cell identity uncertainty propagates into specificity. Mixed or transitional clusters may simultaneously express ligand machinery and receptors, creating apparent self or reciprocal routes. Repeat the analysis at a broader hierarchy and inspect co-detection within individual cells before describing an autocrine loop.

Spatial validation should use route-specific expectations. Synaptic signalling, paracrine diffusion and membrane contact operate on different scales. A single neighbourhood radius cannot validate them all. Record the anatomical compartment and signal class, then choose imaging or spatial measurements that match the proposed biology.

Comparisons with other methods should preserve the same cells, labels and eligible protein interactions. The molecular-ligand feature may have no direct analogue elsewhere, so benchmark its components against targeted chemistry or a transparent gene-set baseline. Do not declare superiority merely because scConnect returns an additional class of edges.

Negative controls can include cell pairs from incompatible compartments, shuffled receptor identities with preserved frequency, molecular ligands lacking required biosynthetic machinery and known pharmacological interactions absent from the tissue. Each control tests a different step. One omnibus randomisation rarely diagnoses every failure mode.

Validation should start at the weakest link. If receptor expression is sparse, confirm receptor protein first. If the ensemble estimate depends on one transporter, measure ligand abundance or flux. If cells are distant, test anatomy before a costly perturbation. This sequence uses the graph to allocate experiments efficiently.

Time-resolved assays are valuable because transcript capacity can precede or follow release. Measure the candidate ligand or proximal receiver event soon after stimulation, then a later gene programme or phenotype. A downstream response present before the proposed source activates challenges the route.

Archive the pharmacology resource and gene-to-ligand model as versioned data, not merely software dependencies. Updates can change eligible receptors or component sets while the specimen stays fixed. Reanalysis should distinguish a knowledge-base change from a biological change.

A calibrated report includes one robust protein route, one robust molecular-ligand route, one specificity-sensitive route and one negative control. Applying the same validation ladder to all four shows where the framework performs well and where its assumptions matter.

The bounded conclusion is practical: scConnect creates a directed, pharmacology-informed shortlist whose specificity and molecular-ligand estimates can guide experiments. Its best use is to convert a large expression dataset into competing, falsifiable routes while keeping the unmeasured molecular steps visible.

Contents · Previous section · Continue to the Science Learning Hub