eduKateSG · Genetics · The Gold Standard of Education
Infer kinship from identity-by-descent segments without mistaking an algorithmic call for a family fact
Choose the explanation you need, or follow all twelve chapters in order.
Full chapter index · Independent practice · Genetics subject guide
Identity-by-descent segments are chromosome stretches that two people are inferred to have inherited from a shared ancestral copy. Their total length, number and sharing state can help distinguish close relationship classes, but the inference depends on genotypes, phase, genetic maps, calling thresholds and population context.
This guide follows the full reasoning chain: from inheritance and IBD0/IBD1/IBD2, through phasing and segment detection, to kinship coefficients, relationship classification and ethical reporting. The examples are fictional. They teach how to audit evidence, not how to identify a real person or reveal an unexpected family relationship.
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
CHAPTER 1 OF 12 · See the mechanism
1. Separate identity by state from identity by descent
Two alleles are identical by state when their observed values match. Identity by descent adds a genealogical claim: the matching copies descend from the same ancestral copy. At a single common marker, matching states provide weak evidence of descent because many unrelated people can carry the same allele.
IBD-segment methods therefore look for extended haplotype sharing across many markers. A long, coherent shared tract is less likely to arise by chance than one matching marker. Even then, the tract is inferred under a model. Marker density, allele frequencies, genotype error and linkage disequilibrium affect what the software sees.
Write the observation and interpretation separately. “The caller reported a 24 cM segment” is an output. “The pair inherited that segment from a recent common ancestor” is the intended biological explanation, qualified by validation. “They are second cousins” is a further classification that may remain ambiguous.
Evidence check: for Separate identity by state from identity by descent, preserve the observation before the interpretation. Record sample, build, genetic map, callable genome 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.
At an autosomal position, a pair can share zero, one or two allele copies IBD. The genome-wide proportions are commonly written k0, k1 and k2. Parent–child pairs are expected to share one copy across nearly the entire autosomal genome, whereas full siblings alternate among all three states because they may inherit neither, one or both corresponding parental haplotypes.
The familiar kinship coefficient can be expressed as phi = k1/4 + k2/2. It is the probability that randomly selected alleles, one from each person at the same autosomal location, are IBD under the model. It is not simply the fraction of markers that match.
Expected proportions describe relationship classes, not exact individuals. Meiosis is random, so relatives at the same degree can share different totals. More distant classes overlap strongly. Preserve k0, k1, k2, total IBD length and segment count rather than collapsing everything into one label.
Evidence check: for Use IBD0, IBD1 and IBD2 as sharing states, preserve the observation before the interpretation. Record IBS, IBD0, IBD1 and IBD2 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 why phase affects segment calls
Diploid genotypes usually show the two alleles at each site without directly saying which allele lies on the chromosome copy inherited from each parent. Phasing estimates those haplotypes. A switch error assigns the haplotype incorrectly beyond a position and can split, shorten or misconnect a true shared segment.
Phase-aware IBD callers gain power by comparing haplotypes, but their accuracy follows the phase quality. Reference-panel fit, sample size, ancestry composition, related individuals, marker density and genotype uncertainty all matter. A phase-free method avoids one dependency but still makes assumptions about continuous sharing and tolerated mismatch.
Record the phasing program, version, reference panel, genetic map and whether relatives were phased jointly. Do not “repair” a result by joining gaps without a declared rule. If different phase seeds or methods change the relationship class, that instability belongs in the conclusion.
Evidence check: for Understand why phase affects segment calls, preserve the observation before the interpretation. Record phasing, panel, caller and threshold contract. 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 callable autosomal genome is 3,400 cM. A fictional pair has 1,720 cM called IBD1 and 620 cM called IBD2 after quality filtering. Then k1 = 1,720/3,400 = 0.506 and k2 = 620/3,400 = 0.182. The estimated kinship coefficient is 0.506/4 + 0.182/2 = 0.218.
That value is nearer a first-degree expectation than a second-degree expectation, but the sharing-state pattern is essential. A parent–child pair should have near-zero IBD2 and almost complete IBD1 coverage. Substantial IBD2 with all three states is more compatible with full siblings, subject to sex-chromosome, sample and data checks.
The numbers are illustrative, not a universal classification boundary. The callable denominator, edge trimming and definition of IBD2 change the estimate. Retain unrounded totals and reconstruct the coefficient from exported segments.
| Quantity | Value | Calculation | Meaning |
|---|---|---|---|
| Callable autosomes | 3,400 cM | Quality-filtered denominator | Not physical genome size |
| IBD1 length | 1,720 cM | One copy shared | k1 = 0.506 |
| IBD2 length | 620 cM | Two copies shared | k2 = 0.182 |
| Kinship coefficient | 0.218 | k1/4 + k2/2 | First-degree range, pattern still needed |
Evidence check: for Calculate a fictional kinship summary, preserve the observation before the interpretation. Record every k value and kinship division. 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.
Recombination breaks ancestral chromosomes across generations. Close relatives usually share longer and more total IBD than distant relatives. Yet two relationships with similar total sharing can produce different segment counts, and the distributions for adjacent distant degrees overlap.
Very short calls are most vulnerable to population linkage disequilibrium, conflation of nearby ancestral tracts and false positives. Long segments can still be fragmented by phasing or genotyping error. A minimum centimorgan threshold therefore changes both the total and the apparent number of ancestors.
Use simulated or empirical reference distributions produced with the same pipeline. Compare total length, IBD2 proportion, segment count and longest segment. Report a set of compatible relationship classes when evidence does not separate them, rather than choosing the most familiar family term.
Evidence check: for Interpret length and count together, preserve the observation before the interpretation. Record length and count as probabilistic 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.
Centimorgans measure recombination probability, while megabases measure physical distance. The conversion varies across chromosomes and populations. A physically long segment in a low-recombination region may be genetically modest; a shorter physical interval can span more genetic distance elsewhere.
Use the genome build and genetic map expected by the software. Coordinate mismatches can shift markers, invert order or create false gaps. Check centromeres, telomeres, low-complexity regions and structural variants before accepting an endpoint.
Boundary uncertainty is often more honest than a single base-pair coordinate. For relationship classification, modest endpoint differences may have little effect; for ancestor tracing or rare-variant follow-up, they may matter greatly. Match reporting precision to the actual task.
Evidence check: for Check genetic maps and boundaries, preserve the observation before the interpretation. Record genetic versus physical boundaries. 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 7 OF 12 · Test what changes
7. Treat genotyping error as directional evidence
A false heterozygous call can break a true IBD tract, while tolerated mismatches can join tracts that should remain separate. Missingness and allele dropout can distort IBD2 especially. Low-quality or degraded samples need their own validation, not thresholds copied from high-quality arrays.
Plot discordance and missingness within called segments. Use Mendelian checks in known trios when available, technical replicates, and concordance across platforms for consequential findings. Sensitivity analysis should vary mismatch allowance and minimum length together because they interact.
Do not select the parameter setting that gives the expected pedigree. Predeclare a primary contract, show reasonable alternatives and explain which evidence supports any revision. The goal is not a neat label; it is a result another analyst can reproduce.
Evidence check: for Treat genotyping error as directional evidence, preserve the observation before the interpretation. Record state pattern versus total sharing. 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.
Population history changes background haplotype sharing. Bottlenecks, isolation and endogamy can increase the number of shared segments among people whose recent paper pedigrees do not connect. Allele-frequency-based coefficients can also be biased when the reference population is poorly matched.
Segment methods reduce some frequency dependence but do not erase demography. A pair may share many short and intermediate segments through multiple distant paths. Algorithms that assume a simple outbred pedigree can assign a relationship that is too close.
Use ancestry-aware reference data, inspect the length spectrum and consider pedigree simulations that reflect the population. State when “equivalent sharing to” is safer than a named relationship. Social identity must not be inferred from an IBD pattern.
Evidence check: for Account for population structure and endogamy, preserve the observation before the interpretation. Record population history and reference fit. 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. Classify relationships with explicit uncertainty
Classification can combine k0, k1, k2, total IBD length, segment count, ages and known pedigree constraints. Each added feature must be documented. Ages may rule out some genealogical arrangements but cannot establish biological kinship by themselves.
Cross-validation should keep families together so close relatives do not leak between training and test sets. Performance measured on one ancestry group or genotyping platform may not transfer. Confusion matrices are more informative than a single accuracy percentage.
For a real report, give the model version, candidate classes, probabilities or intervals if calibrated, and important alternatives. Avoid reporting more precision than the validation supports. “First-degree, with a full-sibling pattern” is stronger and more auditable than an unsupported exact family story.
| Evidence | Primary use | Failure checked | Limit |
|---|---|---|---|
| k0/k1/k2 | Sharing-state pattern | Wrong relationship family | Caller-dependent |
| Segment lengths | Generation signal | Short false positives | Overlapping classes |
| Segment count | Refine distant classes | Conflation and splits | Population history |
| Pedigree and ages | Constrain arrangements | Impossible timelines | May be incomplete |
Evidence check: for Classify relationships with explicit uncertainty, preserve the observation before the interpretation. Record segment, classifier and family uncertainty. 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. Protect consent, privacy and family relationships
Kinship inference can reveal misattributed parentage, donor conception, adoption, previously unknown siblings and population connections. These are not merely technical surprises. Consent should cover relationship discovery, secondary use, data retention and who may receive results.
Separate research, forensic, genealogical and clinical purposes. Their standards, legal authority and consequences differ. An educational workflow cannot substitute for accredited practice, qualified counselling or jurisdiction-specific rules.
Minimise identifiable exports, restrict access and keep an incident plan. Do not contact inferred relatives or expose a family conclusion without the appropriate consent and professional process. Technical confidence does not remove the human duty to handle uncertainty carefully.
Evidence check: for Protect consent, privacy and family relationships, preserve the observation before the interpretation. Record consent, privacy and surprise 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.
Pair A and Pair B both share 1,700 cM in total. Pair A has substantial IBD2 and a mixture of IBD0, IBD1 and IBD2; Pair B has almost no IBD2 and is IBD1 across roughly half the genome. The equal total does not imply the same relationship. A is more compatible with full siblings; B is more compatible with a second-degree pattern, subject to validated thresholds.
Now rephase Pair A with a better matched reference panel. Several apparent IBD2 stretches become adjacent IBD1 tracts and phi falls. Recompute from the revised states and widen the candidate set. The correct response to improved evidence is a narrower claim, not loyalty to the first label.
Evidence check: for Run a changed-context transfer test, preserve the observation before the interpretation. Record the equal-total changed-state 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, purpose, genome build, platform, sample quality, marker filters, phase method, reference panel, genetic map, caller version, thresholds, callable length, segment file, k0/k1/k2, kinship calculation, population reference, sensitivity runs and report wording. Common questions: Does matching DNA prove the exact relationship? No. Is a short segment always real? No. Can equal totals represent different classes? Yes. Does phase error only reduce sharing? It can split, shorten or misassign tracts.
Continue with runs of homozygosity and recent shared ancestry and inbreeding load and purging tests. For methods and limits, read the open IBDkin study, the TPBWT evaluation of phase-aware IBD inference, and the study of IBD segment counts for relationship inference.
Evidence check: for Keep the audit trail and continue, preserve the observation before the interpretation. Record the narrowest kinship class 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.