Why Science? | Single-Cell Hi-C, Sparse Contact Maps and Cell-to-Cell Genome-Folding Evidence

Section 1 of 36

1. Begin with one nucleus, not an average

Single-cell Hi-C asks which contacts can be recovered from an individual nucleus. Population Hi-C smooths millions of conformations into a dense map; a single-cell library keeps cellular identity but captures only a sparse subset of that cell’s possible interactions.

Contents · Next section

Section 2 of 36

2. Did you know sparse can still be informative?

A cell may yield far fewer contacts than an ensemble library, yet chromosome territories, large domains, distance patterns and cell-cycle state can still leave statistical signatures. Sparse does not mean useless; it means the model must respect missing data.

Contents · Previous section · Next section

Section 3 of 36

3. Separate one-cell evidence from a complete structure

Recovered pairs are molecular observations from one processed nucleus. A three-dimensional reconstruction fills many unobserved positions through assumptions and restraints. Display measured contacts and modelled coordinates as different evidence layers.

Contents · Previous section · Next section

Section 4 of 36

4. Keep in-nucleus ligation central

Single-cell protocols move proximity ligation into intact nuclei before individual isolation. This helps preserve contacts and prevents separated libraries from exchanging material. Nuclear integrity and free-DNA controls remain essential because early damage follows every later step.

Contents · Previous section · Next section

Section 5 of 36

5. Understand why population domains can look different per cell

A topologically associating domain in a population map can emerge from recurring but variable contacts across cells. One nucleus need not contain the full average pattern. Single-cell data let researchers ask how often and in what form organisation recurs.

Contents · Previous section · Next section

Section 6 of 36

6. Treat cell cycle as biology and a confounder

Chromosomes reorganise through G1, S phase, G2 and mitosis. A cluster of similar contact maps may reflect cell-cycle position rather than a stable cell identity. Independent cell-cycle measurements and trajectory checks protect interpretation.

Contents · Previous section · Next section

Section 7 of 36

7. Plan cells, depth and replication together

Predefine biological specimens, target cells per specimen, expected contacts per cell, minimum cis fraction and exclusion rules. Thousands of cells from one preparation do not replace independent biological replication when the claim concerns a condition or population.

Contents · Previous section · Next section

Section 8 of 36

8. Crosslink cells consistently

Matched fixation preserves comparable neighbourhoods. Measure viability, cell size and nuclear morphology before and after fixation. A condition that alters fragility can change contact recovery and cell inclusion long before computational clustering.

Contents · Previous section · Next section

Section 9 of 36

9. Digest chromatin inside nuclei

Restriction digestion creates ends for proximity ligation. Report enzyme, fragment geometry and efficiency at several loci. Under-digestion lowers contacts; over-handling can damage nuclei. Cell-type differences in access may require optimisation rather than one universal condition.

Contents · Previous section · Next section

Section 10 of 36

10. Fill ends and mark junctions where the protocol requires

Hi-C chemistries may fill restriction overhangs with biotinylated nucleotides before ligation. Record incorporation and cleanup. The exact chemistry determines which junctions can later be enriched and how artifacts are recognised.

Contents · Previous section · Next section

Section 11 of 36

11. Ligate before isolating individual nuclei

In-nucleus ligation encodes proximity into DNA molecules while cells remain separated physical compartments. Low ligation yield leaves extremely sparse maps; nuclear breakage can create cross-cell background. Test both with bulk-quality controls before single-cell sorting.

Contents · Previous section · Next section

Section 12 of 36

12. Isolate one nucleus with an auditable route

Fluorescence-activated sorting, micromanipulation or another isolation method should include doublet controls, empty wells and plate maps. A doublet can look like one unusually connected cell. Record index, gate, well and processing batch.

Contents · Previous section · Next section

Section 13 of 36

13. Reverse crosslinks and recover tiny DNA amounts

Each nucleus contains limited material. Losses during lysis, purification and transfers reduce contact coverage. Low-retention consumables, consistent volumes and controls matter, but recovery claims should be supported by library complexity rather than assumed from protocol fidelity.

Contents · Previous section · Next section

Section 14 of 36

14. Amplify without turning noise into structure

Library amplification is necessary and potentially distorting. Report cycles, duplicate rate, coverage balance and chimeric products. A cell with many reads may still have few unique contacts if one small set of molecules amplified preferentially.

Contents · Previous section · Next section

Section 15 of 36

15. Practise with an invented single-cell Hi-C table

These fictional cells teach quality review; they are not biological benchmarks.

Contents · Previous section · Next section

Section 16 of 36

16. Demultiplex cell identity before biology

Read barcodes, plate indexes or other identifiers with declared mismatch rules. Exclude ambiguous and cross-contaminated assignments. A barcode correction should never merge two plausible cells merely to raise coverage.

Contents · Previous section · Next section

Section 17 of 36

17. Map pairs with haploid and diploid limits in mind

State genome build, aligner, mapping threshold and repeat handling. Without phased variants, homologous chromosomes may be indistinguishable. A contact assigned to chromosome 1 is not automatically assigned to the maternal or paternal copy.

Contents · Previous section · Next section

Section 18 of 36

18. Filter invalid pairs and duplicates per cell

Count dangling ends, self-circles, same-fragment pairs, duplicates, low-quality mappings and cis/trans contacts for every cell. Per-cell quality distributions reveal damaged nuclei and batch shifts that disappear in pooled summaries.

Contents · Previous section · Next section

Section 19 of 36

19. Use distance curves as a quality and state signal

Cis-contact probability normally declines with genomic distance, while chromosome compaction changes through the cell cycle. Compare cells on matched unique-contact counts and inspect technical covariates before treating curve differences as new biology.

Contents · Previous section · Next section

Section 20 of 36

20. Cluster cells without letting depth lead

Dimensionality reduction and clustering can separate cells by contact pattern, but depth, cis fraction and batch may dominate. Downsample, regress or stratify technical variables and confirm clusters across specimens.

Contents · Previous section · Next section

Section 21 of 36

21. Reconstruct structures with explicit restraints

A model can place genomic bins in three dimensions so observed contacts are nearby and chromosome geometry is plausible. Report objective function, resolution, ensemble of solutions and uncertainty. One attractive rendering is not a uniquely measured nucleus.

Contents · Previous section · Next section

Section 22 of 36

22. Challenge sparsity-driven absence

If a contact is missing from one cell, it may simply not have been sampled. Avoid binary present-versus-absent claims at low coverage. Aggregate defined features, model detection probability and use many cells to estimate recurrence.

Contents · Previous section · Next section

Section 23 of 36

23. Challenge doublets and mixed nuclei

Two nuclei entering one well can inflate trans contacts and create impossible ploidy. Sorting gates, species mixtures, allele balance and contact-pattern diagnostics can reveal doublets. Exclude them under rules set before biological clustering.

Contents · Previous section · Next section

Section 24 of 36

24. Challenge amplification artifacts

Whole-genome or library amplification can create uneven coverage and chimeras. Use molecule-level duplicate information, negative controls and coverage plots. A dense local cluster supported by one amplified family is not a chromosome domain.

Contents · Previous section · Next section

Section 25 of 36

25. Challenge cell-cycle stories

A smooth computational trajectory may align with replication timing and contact distance, but ordering is an inference. Validate with DNA content, replication markers or synchronised references and check that batch does not mimic progression.

Contents · Previous section · Next section

Section 26 of 36

26. Report variability with specimen-level support

Show how feature frequencies vary across cells within each independent specimen. Hundreds of cells improve the estimate of within-sample heterogeneity; they do not convert one donor or culture into broad population replication.

Contents · Previous section · Next section

Section 27 of 36

27. Validate structures with imaging or perturbation

Fluorescence imaging can test distances, territories and nuclear position. Perturbation can test whether a factor affects a recurrent feature. Use orthogonal evidence to challenge the model, especially when fine coordinates were imputed from sparse contacts.

Contents · Previous section · Next section

Section 28 of 36

28. Learn single-cell Hi-C with sparse dot cards

Give each imaginary cell a different small sample of dots from a hidden pattern. Pooling reveals the average; keeping cards separate reveals variability. The exercise shows why absence from one card is weak evidence and repeated structure across cards is stronger.

Contents · Previous section · Next section

Section 29 of 36

29. Build Primary Science process skills

Students can compare equal-sized samples, count recurring categories and explain why more observations improve confidence. The paper model teaches classification, variation, repeated trials and fair comparison without any biological material.

Contents · Previous section · Next section

Section 30 of 36

30. Prepare for PSLE Science reasoning

Ask which cell fails quality checks, why a doublet control is needed and whether missing evidence proves absence. These questions practise data interpretation and claim limits in a modern context.

Contents · Previous section · Next section

Section 31 of 36

31. Extend into Secondary and O-Level Science

Single-cell Hi-C connects chromosomes, cell cycle, enzymes, probability, sampling, graphs and models. Students can separate measured pairs from inferred structure and discuss how technical variation competes with biological variation.

Contents · Previous section · Next section

Section 32 of 36

32. Use the topic for school choices

Check current official school pages for practical science, computing, research and data opportunities. Ask how projects teach sampling and uncertainty. Avoid assuming a specialist programme guarantees admission, fit or a career result.

Contents · Previous section · Next section

Section 33 of 36

33. See the career ecosystem without promises

Work may involve cell biology, flow sorting, sequencing, statistics, software, modelling and imaging. Qualifications differ by role. Teams need both careful experimentalists and analysts willing to expose uncertainty.

Contents · Previous section · Next section

Section 34 of 36

34. Use questions for science tuition and enrichment

Ask why ligation occurs before isolation, why maps are sparse and how a doublet changes the result. These questions turn single-cell genomics into reasoning practice across Primary Science, PSLE Science, Secondary Science and O-Level Science.

Contents · Previous section · Next section

Section 35 of 36

35. Did you know variation can be the result?

Population experiments often treat cell-to-cell differences as noise. Single-cell Hi-C can make organised variability measurable. The scientific task is to separate meaningful heterogeneity from sampling, amplification, batch and cell-cycle effects.

Contents · Previous section · Next section

Section 36 of 36

36. Conclude with a single-cell Hi-C evidence checklist

Ask whether nuclei stayed intact; isolation excluded doublets; unique contacts and cis fractions were adequate; amplification and mapping were audited per cell; sparsity and cell cycle were modelled; specimens replicated the pattern; and structures were challenged by independent evidence.

A strong single-cell Hi-C project starts with a sampling memo. Name the biological population, independent specimens, cell-cycle range, nuclei isolation, per-cell library method, target cell count, expected unique contacts and primary feature tests. Predefine doublet, cis-fraction, duplicate and coverage exclusions. Separate goals that need many cells from goals that need deep cells. A larger cell count estimates heterogeneity; deeper individual libraries support finer structure, and one cannot automatically replace the other.

Quality control must stay attached to each cell. Keep sorting gate, plate position, processing batch, read count, unique contacts, cis/trans balance, distance curve, duplicate rate, chromosome coverage and allele balance. Visualise distributions before choosing thresholds. If excluded cells cluster by treatment or specimen, the filter may remove biology as well as failures. Report that asymmetry and test whether conclusions change under reasonable inclusion rules.

Sparse maps need coverage-aware statistics. Absence of one pair from one cell has little meaning; recurrence of a domain-scale feature across many adequately sampled cells can be informative. Use matched downsampling, detection models and specimen-aware summaries. Avoid smoothing every cell until it resembles the population average. If imputation is used, publish both raw and imputed results and state which conclusions depend on filled-in contacts.

Cell-cycle inference should not become circular. Contact maps can position cells along a progression, but the same features should not define and validate the trajectory. Use DNA content, replication signals, microscopy or an independent reference where possible. Show technical covariates across the inferred path. A gradual shift in read depth or cis fraction can imitate a biological continuum if quality control is separated from the final figure.

Structural modelling requires ensembles and uncertainty. Report resolution, restraints, polymer assumptions, nuclear boundary, optimisation, initialisations and fit diagnostics. Compare models built from downsampled contacts. Display positional uncertainty and distinguish observed pair distances from inferred coordinates. One rendering can help readers see a hypothesis, but multiple compatible solutions should remain part of the evidence story.

Figures should begin with the cell-quality landscape: sorting, contacts, duplicates, cis fraction, distance curves, chromosome coverage, batch and specimen. Next show recurrence of domains or territories before the most dramatic 3D rendering. Use consistent scales and include failed cells as grey quality references when appropriate. Label whether each panel is one cell, an aggregate, an embedding or a model.

Stewardship includes consent, specimen and cell metadata, fixation, enzyme lots, sorter gates, plate maps, indexes, raw reads, per-cell pairs, exclusion tables, embeddings, model ensembles, code and software environments. Single-cell human data can expose sequence variants and require controlled access. Laboratory safety covers biological material, fixatives, flow cytometry, enzymes, heat and amplified DNA; computational auditability protects later biological decisions.

The next experiment should address the largest uncertainty. Add cells when feature frequencies are imprecise, deepen libraries when individual structures are too sparse, tighten sorting when doublets rise, rebuild poor batches, or add imaging when geometry matters. Perturbation can test whether a recurrent folding feature depends on a factor. State in advance what result would strengthen, narrow or contradict the proposed explanation.

For students, the method offers an optimistic lesson about variation. If every learner receives only a few dots from a hidden picture, pooling reveals the average while keeping sheets separate reveals diversity. Both views are useful. The exercise connects observation, sampling, classification and uncertainty across Primary Science, PSLE Science, Secondary Science and O-Level Science without suggesting a sparse sheet is a failed experiment.

A compact reproducibility exercise can make sparsity tangible. Give two analysts the same filtered cells but require them to choose clustering resolution, distance normalisation and minimum coverage before seeing labels. Compare the resulting groups, then repeat after downsampling richer cells to the same contact count as poorer ones. Stable biological structure should not depend entirely on which cells had the most reads. Recording these choices beside the figures turns apparent cell-to-cell diversity into a testable result and teaches that a single-cell atlas is assembled from many uncertain observations, not delivered fully formed by the instrument.

Replicate structure must remain biological, not merely cellular. Hundreds of nuclei from one specimen do not substitute for independent specimens when a conclusion concerns people, tissues or treatments. Summaries should therefore show cells nested inside their donors or experiments, with uncertainty calculated at the level that matches the claim. A cluster that exists in one preparation may be a real local state, a batch effect or both. Balanced sampling, held-out specimens and sensitivity analyses give readers a way to tell. This distinction is valuable far beyond genomics because it separates repeated measurements from genuinely independent evidence.

Publish the cell-by-feature matrix and the precise cell identifiers behind every summary so that the aggregation can be audited and reproduced.

Contents · Previous section · Continue to the Science Learning Hub