eduKateSG · Genetics · The Gold Standard of Education
Read chromosome painting as a model of haplotype copying, not a literal map of identity
Choose the explanation you need, or follow all twelve chapters in order.
Full chapter index · Independent practice · Genetics subject guide
Chromosome painting methods represent each recipient haplotype as a mosaic copied from reference haplotypes. The pattern can reveal fine-scale population structure, admixture and shared ancestry, but every colour depends on the donor panel, phasing, recombination map and model. A painted segment is an inference, not a physical label embedded in DNA.
This guide connects haplotypes, copying models, ancestry tracts and time since admixture. It keeps three questions separate: which reference haplotype best explains a segment, which historical population process could create the pattern, and what the result cannot say about a person’s social identity.
Continue from The Gold Standard of Genetics within The Gold Standard of Education: learning that transfers into life and the How X Works learning library.
Full chapter index
See the mechanism · Chapters 1–2
Build the worked model · Chapters 3–5
Test what changes · Chapters 6–8
Check evidence and limits · Chapters 9–10
Practise and continue · Chapters 11–12
Chromosome painting asks which reference haplotypes provide the best model for successive parts of a recipient chromosome. Identity-by-descent analysis asks whether two copies descend from a shared ancestral copy within a modelled timescale. Local ancestry analysis assigns chromosome regions to ancestry-source categories. These tasks overlap, but they are not interchangeable.
A copying chunk need not be a recently inherited IBD segment from the named donor. The donor can stand in for a broader pool of related haplotypes. Likewise, a local ancestry label depends on available source samples and classification rules.
Write the output precisely: “this interval was most strongly copied from reference group A under this panel” is narrower than “this interval came from population A.” The first is an auditable model statement; the second may overstate history and identity.
Evidence check: for Distinguish painting, IBD and local ancestry, preserve the observation before the interpretation. Record recipient, donor panel, build, markers and task. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
CHAPTER 2 OF 12 · See the mechanism
2. Start with phased haplotypes and recombination
Diploid genotypes must usually be phased so alleles are arranged into estimated chromosome copies. Switch errors can break the copying path or move a segment between haplotypes. Rare variants, sample size and reference-panel composition affect phase quality.
Recombination makes chromosomes mosaics across generations. Painting algorithms use a genetic map or recombination model to balance persistence of one donor against switches to another. Centimorgans therefore matter differently from physical megabases.
Record genome build, marker set, phasing method, map, missingness and whether close relatives were present. If the scientific question concerns very short chunks, phase and map uncertainty may dominate the historical signal.
Evidence check: for Start with phased haplotypes and recombination, preserve the observation before the interpretation. Record painting, IBD and local ancestry separately. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
CHAPTER 3 OF 12 · Build the worked model
3. Understand the recipient–donor copying model
For each recipient position, the model considers reference haplotypes as possible donors. It assigns copying probabilities or a most likely donor and allows switches along the chromosome. Emission parameters accommodate mutation and genotyping differences; transition parameters reflect recombination and effective population assumptions.
The result can be summarised as chunk counts, total copied length or a coancestry matrix. Those summaries describe resemblance within the supplied model. They do not mean every reference individual is a genealogical ancestor.
Inspect both directions of comparison where the design permits. A group that copies strongly from another need not receive the same pattern in return because sample sizes, diversity and drift differ. Apparent asymmetry can be historically informative, but it can also be created by the donor set. Balance and resampling tests help separate those explanations.
Decide whether recipients may copy from their own group and whether close relatives should be excluded. Self-copying can obscure between-group structure; removing too many donors can make a population appear artificially distinct. The rule must match the question.
Evidence check: for Understand the recipient–donor copying model, preserve the observation before the interpretation. Record phase, map and copying parameters. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Suppose a recipient has 3,000 cM of callable autosomal sequence. A painting run assigns 1,650 cM to reference cluster A, 900 cM to cluster B, 300 cM to cluster C and 150 cM remains weak or ambiguous. The nominal shares are 55%, 30%, 10% and 5%.
These are copying proportions under this reference panel. They are not automatically literal ancestry percentages. Cluster A may be overrepresented, internally diverse or genetically close to an unsampled source. The 5% ambiguous component should not be forced into the largest colour.
Preserve chromosome-level chunks and posterior support. Two people with the same global proportions can have different tract lengths and histories. Global totals are a useful summary, not the full evidence.
| Copying source | Length | Nominal share | Interpretation |
|---|---|---|---|
| Cluster A | 1,650 cM | 55% | Panel-dependent copying |
| Cluster B | 900 cM | 30% | Not literal identity |
| Cluster C | 300 cM | 10% | Reference surrogate |
| Ambiguous | 150 cM | 5% | Do not force assignment |
Evidence check: for Work a fictional copying summary, preserve the observation before the interpretation. Record every length, share and ambiguous component. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
After admixture, recombination tends to break source-associated ancestry tracts across generations. A recent mixture event usually leaves longer tracts than an older event, all else equal. Models can use the distribution of chunk lengths or ancestry covariance to estimate dates.
The clock is probabilistic. Continuous migration, multiple pulses, assortative mating, selection and map error can produce departures from a single-pulse model. Short chunks are also harder to assign reliably.
Report the demographic model with the date. “Twenty generations under a single-pulse model” is not equivalent to a known historical event twenty generations ago. Compare alternative histories and show uncertainty intervals.
Evidence check: for Let tract length inform time cautiously, preserve the observation before the interpretation. Record tract length as model-dependent time evidence. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
CHAPTER 6 OF 12 · Test what changes
6. Treat the reference panel as part of the result
A reference panel defines the alternatives the algorithm can recognise. If the true source is absent, the model may copy from a related available group. Small panels underrepresent within-group diversity and increase assignment error.
Panel labels often combine geography, recruitment and researcher choices. They should not be treated as natural biological boundaries. Document sampling place and time, inclusion rules, relatedness filters and sample counts.
Run leave-one-group-out and balanced-panel checks. If a conclusion disappears when donor counts are equalised or a neighbouring group is added, state that dependence. Better representation can change both proportions and the historical story.
Evidence check: for Treat the reference panel as part of the result, preserve the observation before the interpretation. Record panel labels, counts and source gaps. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Painting can reveal clusters because members share copying profiles. Clustering is a statistical description of the sample, not proof of isolated or pure populations. Geographic gradients and recent migration can create continuous structure that a tree divides into boxes.
Admixture inference needs a comparison between the target pattern and plausible source surrogates. Related sources can be hard to distinguish. A fitted mixture may explain the data without being the only historical explanation.
Use principal components, allele-frequency methods, historical information and explicit demographic simulations as complementary evidence. Agreement across methods can strengthen an inference; disagreement can reveal which assumptions require attention.
Evidence check: for Separate population structure from admixture, preserve the observation before the interpretation. Record structure versus admixture histories. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Posterior probabilities often change gradually near tract boundaries. A single switch coordinate can imply more precision than the data support. Report uncertainty bands or window-level probabilities when downstream conclusions depend on endpoints.
Low marker density, centromeres, structural variation and low-complexity regions can cause uncertain or spurious switches. Masking rules should be fixed before comparing groups.
Avoid smoothing only until the picture looks historically plausible. Declare minimum tract length and confidence thresholds, then show sensitivity. Ambiguous sequence is information about uncertainty, not clutter to be painted away.
Evidence check: for Audit boundaries and ambiguous windows, preserve the observation before the interpretation. Record boundary support and masked regions. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
CHAPTER 9 OF 12 · Check evidence and limits
9. Validate with simulations and held-out samples
Simulate haplotypes under demographic histories resembling the task, then pass them through phasing and painting. Measure assignment accuracy, boundary error and date recovery. Simple simulations that omit reference imbalance or continuous gene flow can overstate performance.
Hold out individuals or populations and test whether the method recognises related structure without seeing the label. Replicate with another genotyping platform or model where possible.
Keep family members in the same validation partition. Otherwise, near-identical haplotypes can leak into training and make ancestry assignment appear easier than it is. Report confusion among neighbouring sources, not only overall accuracy.
| Check | Primary question | Failure detected | Needed record |
|---|---|---|---|
| Simulation | Can history be recovered? | Model overconfidence | Generative settings |
| Held-out donor | Does label generalise? | Panel memorisation | Split rule |
| Balanced panel | Are shares stable? | Donor-count bias | Sample counts |
| Alternative method | Does structure persist? | Method-specific pattern | Comparable inputs |
Evidence check: for Validate with simulations and held-out samples, preserve the observation before the interpretation. Record simulation, holdout and balance checks. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
CHAPTER 10 OF 12 · Check evidence and limits
10. State biological, historical and ethical limits
Genetic ancestry is multidimensional and continuous. It is not the same as race, ethnicity, nationality, culture, kinship law or personal belonging. A reference label should never become a judgement about authenticity.
Present-day samples are imperfect surrogates for past populations. Migration changes every population, and some histories are poorly represented because sampling and archives are unequal. Avoid language of purity or replacement when evidence supports mixture and continuity.
Consent and community engagement matter when population history is sensitive. Protect individual-level haplotypes and consider how findings may affect groups. Technical resolution does not grant authority to define people from outside their own social histories.
Evidence check: for State biological, historical and ethical limits, preserve the observation before the interpretation. Record identity, consent and community limits. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Run 1 uses equal donor counts from A, B and C and assigns 55%, 30%, 10% plus 5% ambiguous. Run 2 triples A’s donor count without changing recipients and A rises to 64%. Do not report a biological shift. The changed reference opportunity altered the copying competition.
Now add a previously unsampled neighbouring source D. Much of the former B component becomes D with higher posterior support, while long A tracts remain stable. Narrow the conclusion: A sharing is robust to this panel change; the B/D distinction was not.
Evidence check: for Run a changed-context transfer test, preserve the observation before the interpretation. Record the expanded-panel changed case. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.
Archive consent, sample and donor definitions, genome build, markers, phasing, genetic map, relatedness filters, copying parameters, callable length, chunk files, posterior support, ambiguous regions, panel-balance tests, simulations, demographic models and reporting language. Common questions: Is a colour a literal ancestry label? No. Are donors necessarily ancestors? No. Do longer tracts usually suggest more recent sharing? Yes, under stated models. Can adding a reference change the picture? Yes.
Continue with identity-by-descent segments and kinship inference, effective population size and coalescent times, and migration, clines and local adaptation. For the method, read the original chromosome-painting population-structure paper and the open WINC local-ancestry study.
Evidence check: for Keep the evidence route and continue, preserve the observation before the interpretation. Record the narrowest ancestry statement supported. Then state the inference, one credible alternative, and the additional observation that would distinguish them. This compact audit keeps a fluent explanation tied to evidence that another reader can inspect.
Changed-context transfer: repeat this chapter’s decision after changing one consequential condition while holding the others fixed. Calculate or predict what should change, name the threshold or mechanism, and compare with the original case. If the conclusion survives, state the range tested; if it reverses, explain why. Transfer is demonstrated by the changed case, not asserted from the first example.
Practical decision record: write one sentence for what is known, one for what remains uncertain, and one for the next safe action. Include the evidence owner, the latest useful decision time, and the condition that would stop or reverse the action. This turns a technical result into a reviewable choice without pretending uncertainty has disappeared.