Contents
11 — Beyond Mendel Part 2 — Transmission genetics

11 — Beyond Mendel

Before this: Ch 10 · Ch 08 · Time: ~45 min

What you'll be able to do

The core idea

Mendel's model has two layers, and they are almost always taught as one.

The first layer is a transmission rule: segregate, one of your two copies goes into each gamete, and loci on different assort independently. This is a statement about , and it is essentially exact.

The second layer is a -to- map: AA and Aa look like this, aa looks like that. This is a statement about biochemistry, and it is a crude approximation that happens to hold for the seven characters Mendel picked.

Everything in this chapter perturbs the second layer. Nothing in this chapter perturbs the first.

That distinction does real work. When a dihybrid cross comes out 9:3:4 rather than 9:3:3:1, no allele failed to segregate and no gamete was mis-made. The same sixteen genotype classes were produced at the same frequencies. What changed is that the rendering function collapsed two of the phenotypic classes into one. The denominators are untouched — which is precisely why the modified ratios are still sixteenths.

Every "exception to Mendel" in this chapter is a change to the rendering function, not to the protocol. is the protocol. The genotype-to-phenotype map is application logic, and application logic is where all the interesting failures live.


1. Dominance is a property of the phenotype, not of the allele

Start by asking the question nobody asks: why is the default?

Most loss-of-function are recessive. The heterozygote has one working copy of the gene, makes roughly half the normal amount of enzyme, and looks entirely normal. Textbooks assert this. It follows from kinetics.

Consider a metabolic pathway carrying flux J, with one step catalysed by an enzyme present at activity E. Flux does not rise linearly with E forever — it saturates, because as you add more of one enzyme, control passes to the other steps. The curve J(E) is concave, and in a wild-type organism most enzymes sit well out on the flat part, where dJ/dE ≈ 0. Halve E and you slide back along a plateau: flux barely moves. This is the Kacser–Burns argument, and it explains recessivity as an emergent consequence of network kinetics rather than as a property that individual alleles possess.

So dominance is not something an allele does to its partner. Nothing is overpowered. The heterozygote simply lands close to one homozygote on whatever axis you happen to be measuring.

The exceptions tell you the mechanism:

Heterozygote is not normal becauseMechanismExample
50% of product is genuinely not enough — this locus sits on the steep part of the curvePAX6 (aniridia), NF1, many and structural proteins
negativeThe broken product poisons the good one, typically by joining a multimer and jamming itCollagen triple helices in osteogenesis imperfecta
The variant product does something new or does the old thing constitutivelyFGFR3 in achondroplasia
Ectopic expressionNormal product, wrong place or wrong timeMouse Ay, §4

Dosage-sensitive genes are a minority, which is why the recessive default holds. But note what this means: dominance is a claim about the shape of a dose–response curve, and different read out different regions of that curve. Which is the whole content of the next section.

2. Incomplete dominance versus codominance

These are conflated constantly, and the distinction is sharp.

: the heterozygote is intermediate on a single measured scale. One , sitting between the two homozygotes. In snapdragons and four-o'clocks, a red-flowered homozygote crossed to a white gives pink F1, and the F2 is 1 red : 2 pink : 1 white — the phenotypic ratio equals the genotypic ratio, because the map became one-to-one. Pink is not a mixture of red and white flowers; it is a flower with about half the anthocyanin.

: both allelic products are present and separately detectable in the heterozygote. Not an average — a superposition. In the MN blood group, the GYPA encode two forms of glycophorin A differing by two amino acids; an MN heterozygote carries both antigens on every red cell. The F2 ratio is also 1:2:1, but the heterozygote's phenotype is "M and N", not "halfway between M and N".

Incomplete dominanceCodominance
Heterozygote showsone intermediate phenotypeboth parental phenotypes simultaneously
Underlying reasonproduct is dosage-dependent below saturationboth products are made and both are visible
Discriminating questioncan you see the two homozygote phenotypes side by side?
F2 ratio1:2:11:2:1

Now the point that makes the distinction useful rather than pedantic. Take sickle-cell: a single base change in HBB substituting valine for glutamate at the sixth residue of the mature β-globin chain.

AssayWhat the HbA/HbS heterozygote showsDominance verdict
Clinical diseaseessentially healthyHbS is
Red cells under severe hypoxiasome cells sickleHbS is incompletely
Protein electrophoresistwo bands, HbA and HbS, both presentHbS is codominant
Resistance to P. falciparum malariaprotectedHbS is dominant

One allele, one molecular lesion, four dominance verdicts. Dominance is not a property you can look up; it is a property of the (allele, phenotype, assay) triple. The closer your assay gets to the gene product, the more codominant everything becomes — which is why molecular genetics rarely uses the vocabulary at all.

3. Multiple alleles: ABO worked properly

A individual carries at most two . A population can carry any number. Nothing in Mendel's model constrains this, and the classical three-allele ABO system shows two different dominance relationships living inside one locus.

The ABO gene sits at 9q34.2 and encodes a glycosyltransferase — an enzyme that attaches a sugar to a precursor carbohydrate (the H antigen) already sitting on the red-cell surface. Carbohydrate identity tags, exactly as flagged in Ch 01.

Allele (classical)ModernEnzymeSugar addedAntigen
IAABO*Aα-1,3-N-acetylgalactosaminyltransferaseN-acetylgalactosamineA
IBABO*Bα-1,3-galactosyltransferasegalactoseB
iABO*Onone — c.261delG frameshifts to a premature stopnothingH only

Two details make the genetics fall out rather than needing to be learnt:

Now the dominance relationships are forced. Enzymes act in trans on a shared pool of substrate, and their products accumulate independently — activity is additive, not blended:

I^A I^A  →  A antigen only              phenotype A
I^A i    →  A antigen only              phenotype A     (i contributes nothing)
I^B I^B  →  B antigen only              phenotype B
I^B i    →  B antigen only              phenotype B
I^A I^B  →  A antigen AND B antigen     phenotype AB    (both enzymes work)
i i      →  H antigen unmodified        phenotype O

IA and IB are codominant with each other and both are to i — and neither fact required a rule, only the observation that a broken enzyme makes nothing and two working enzymes both make something.

A cross worth doing, because it surprises people: IAi × IBi gives 1 AB : 1 A : 1 B : 1 O. Two parents produce children of all four blood groups — including two, AB and O, that match neither parent. is completely ordinary; the four-way split comes from the map.

(The clinical system has three classical alleles. Sequencing the locus has catalogued hundreds of variants, including weak-A subgroups that behave as intermediates — the three-allele model is an abstraction that survives because it is clinically sufficient.)

4. Lethal alleles and the 2:1 ratio

Cuénot, 1905. Yellow mice never breed true. Yellow × yellow always gives yellow and agouti offspring, in a ratio near 2:1, and litters are about a quarter smaller than expected.

The 2:1 is the tell. A 3:1 with the aa class deleted is 3:1 minus nothing; a 1:2:1 with the homozygous- class deleted is exactly 2:1.

A^y/A  ×  A^y/A
                 1 A^y/A^y      dies before birth
                 2 A^y/A        yellow          }  2 : 1
                 1 A/A          agouti          }

The molecular story is better than the ratio. Ay is a ~170 kb deletion at the mouse agouti locus. It removes the coding sequence of the neighbouring gene Raly while leaving Raly's and non-coding first exon intact — and that orphaned promoter now drives agouti expression ubiquitously instead of in its normal restricted pattern. Two consequences from one deletion:

So Ay is dominant for coat colour and recessive for lethality, simultaneously. If you were still carrying the intuition that dominance is a property of an , this kills it. Dominance is indexed by , and this allele has two.

Human counterpart: achondroplasia (FGFR3). Heterozygotes have the skeletal phenotype; homozygotes die perinatally. Two affected parents expect 2 affected : 1 unaffected among surviving children — the same signature, for the same structural reason.

Distinguishing an observed 2:1 from 3:1 is a hypothesis test with a specific alternative, not an eyeball judgement; that is Ch 12.

5. Pleiotropy

Ay already showed it: one lesion, many . That is , and it is the normal case rather than a curiosity.

Phenylketonuria is the cleanest illustration. A single enzyme, phenylalanine hydroxylase (PAH), fails. Downstream: phenylalanine accumulates and damages the developing brain; tyrosine becomes scarce, so melanin synthesis falls and patients are pale and fair-haired; a phenylalanine derivative gives urine a musty odour; eczema is common. Four apparently unrelated clinical findings, one blocked reaction. Pleiotropy usually means the product is used in more than one context, or one lesion propagates through a network.

The genomics consequence, which will matter repeatedly: because pleiotropy is ubiquitous, "the gene for X" is nearly always a category error, and loci turn up shared across statistically unrelated traits in association studies (Ch 51).

6. Epistasis: pathways generate ratios

is the phenomenon that the phenotypic effect of one locus depends on the at another. Classically (Bateson) it means masking: locus A hides what locus B is doing.

A warning before the ratios. Fisher's statistical epistasis — a non-zero interaction term when you regress on genotypes — is a different object with the same name. A biochemically obvious masking relationship can produce no interaction term under one scale and a large one under another. The two senses coincide only sometimes. Ch 30 takes this apart; for now, keep the words separate in your head.

The derivation you should do instead of memorising

A dihybrid self gives four genotype classes at 9 : 3 : 3 : 1 — namely A_B_, A_bb, aaB_, aabb. Treat "is locus A functional?" and "is locus B functional?" as two binary inputs, weighted 3/4 and 1/4. Every modified ratio is a function from those two inputs onto a smaller output set. There are only so many such functions, the pathway topology picks one, and the ratio follows by summing sixteenths. You never need to remember a ratio again.

graph LR subgraph S1["Sequential — 9:3:4"] P1[precursor] -->|"gene E<br/>MC1R"| I1[eumelanin] -->|"gene B<br/>TYRP1"| O1[black] I1 -.->|"bb"| O2[brown] P1 -.->|"ee — no eumelanin at all"| O3[yellow] end subgraph S2["AND gate — 9:7"] P2[colourless] -->|"gene C"| I2[intermediate] -->|"gene P"| O4[purple] end subgraph S3["OR gate — 15:1"] P3[substrate] -->|"gene A1"| O5[product] P3 -->|"gene A2 — paralogue,<br/>same reaction"| O5 end subgraph S4["Inhibitor — 12:3:1"] P4[precursor] -->|"gene Y"| O6[yellow pigment] W["W — dominant<br/>inhibitor"] -.->|blocks| P4 end

9:3:4 — epistasis. A linear pathway where an upstream step is required for anything at all, and a downstream step only chooses between outcomes. Labrador coat colour: MC1R (the E locus) must signal for eumelanin to be made; TYRP1 (the B locus) determines whether that eumelanin is black or brown.

GenotypeFractionPhenotype
B_E_9/16black
bbE_3/16chocolate
B_ee3/16yellow
bbee1/16yellow

ee masks the B locus entirely: 9 black : 3 chocolate : 4 yellow. The masking is homozygous-recessive, hence "recessive epistasis". A detail that proves the mechanism: the B locus is not silent in ee dogs — it still acts in skin, so a bb ee dog is yellow with a liver-coloured nose. Masking is tissue-specific, because pathways are.

12:3:1 — epistasis. One locus produces a dominant inhibitor. In summer squash, W_ blocks pigment deposition regardless of the Y locus; wwY_ is yellow; wwyy is green. Classes 9 and 3 merge: 12 white : 3 yellow : 1 green.

9:7 — recessive epistasis. A two-step linear pathway in which failure at either step gives the identical null. Bateson and Punnett's sweet peas: genes C and P are both required for purple, and every other genotype is white. An AND gate. Classes 3, 3 and 1 merge: 9 purple : 7 white.

15:1 — duplicate dominant epistasis. Two paralogous genes catalysing the same reaction; either alone suffices. An OR gate, and the usual cause is gene duplication (Ch 35). Shepherd's-purse capsule shape: 15 triangular : 1 ovoid.

Summary, with the rest of the family:

RatioLogic on (A functional, B functional)TopologyClassic example
9:3:3:1independent outputsunrelated pathwaysMendel's dihybrid
9:3:4B readable only if Asequential, upstream gates all outputLabrador coat colour
12:3:1dominant at A blocksinhibitor acting on the pathwaysummer squash fruit colour
9:7A AND Btwo required steps in seriessweet pea flower colour
15:1A OR Bredundant paraloguesshepherd's-purse capsule
13:3inhibitor plus a requirementdominant suppressor over a needed genepoultry plumage colour
9:6:1additive, both contributeduplicate genes with cumulative effectsquash fruit shape

Every entry sums to 16. That is not a coincidence — it is the point.

7. Complementation testing

You run a and recover twenty independent white-flowered mutants. How many genes did you hit?

The 9:7 result answers it. Cross two homozygous mutants and look at the F1, which is a trans heterozygote carrying one on each homologue:

different genes:   m1 +      each locus has one working copy    →  F1 WILD TYPE
                   ── ──     enough product from both           →  they COMPLEMENT
                   +  m2

same gene:         m1 +      no working copy of this gene       →  F1 MUTANT
                   ── ──     nothing to make product from       →  they FAIL to complement
                   m2 +

Partition the mutants by this test and you get complementation groups, which are genes. This was the operational definition of a gene for most of the twentieth century — Benzer's cistron — and it is still how a screen is triaged before anything is sequenced (Ch 37).

Three ways it lies, all of which you will meet:

  1. It is only valid for loss-of-function . A mutation makes the F1 mutant whatever it is crossed to, so it appears to be in every gene. Test dominance first.
  2. Intragenic complementation. Two mutations in the same gene can complement if the product is a multimer and defective subunits can compensate for each other's damaged domains. Rare, but it manufactures a false "different genes".
  3. Non-allelic non-complementation. Two mutations in different genes can fail to complement, typically when the products sit in the same complex and both are dosage-sensitive. This manufactures a false "same gene", and is one of the more useful accidents in genetics — it identifies physically interacting partners.

The formal control is the cis-trans test: put both mutations on the same homologue (m1 m2 / + +). If they are in one gene, the cis configuration leaves one intact copy and is wild-type, while trans is mutant. A difference between cis and trans is the evidence; the trans test alone is a shortcut.

8. Penetrance and expressivity

Two distinct quantities, routinely swapped.

is the proportion of individuals with a who show any of the associated . It is a probability defined over a population, not a property of a person: P(affected | genotype).

is how the phenotype presents among those who show it — severity, which features, which tissues. It is a distribution over the affected.

For a statistician the compact statement is: model phenotype severity as a zero-inflated distribution. Penetrance is one minus the zero-inflation probability. Expressivity is the shape of the non-zero component. They are separately estimable and can move independently.

ConditionPenetranceExpressivity
Postaxial polydactylyincomplete — obligate carriers with normal hands, so the trait skips generationsfairly uniform when present
Neurofibromatosis type 1 (NF1)essentially complete by adulthoodextremely variable — from a handful of café-au-lait macules to disfiguring tumours, and differing between identical twins
Retinoblastoma (RB1)~90% for classic null ; distinct low-penetrance alleles existvariable — unilateral vs bilateral, number of foci
Waardenburg syndromehighvariable — deafness, white forelock, heterochromia, in any combination

Penetrance is a function of age, not a scalar. For anything with adult onset, quoting one number is meaningless; the correct object is a cumulative-incidence curve. In Huntington disease the HTT CAG repeat length sets it explicitly: ≤26 repeats normal, 27–35 intermediate (not disease-causing but unstable in transmission), 36–39 reduced penetrance, ≥40 full penetrance given a normal lifespan. For BRCA1, prospective cohort data give cumulative breast-cancer risk to age 80 of about 72% (95% CI 65–79) and ovarian-cancer risk about 44% (36–53); for BRCA2, about 69% (61–77) and 17% (11–25) respectively (Kuchenbaecker et al., JAMA 2017).

And one methodological trap that matters clinically: penetrance estimated from families ascertained because they contained many affected members is biased upward, because you conditioned on the outcome. Estimates from unselected population cohorts and biobanks come out systematically lower for the same variant. This is a live issue in returning incidental findings (Ch 55).

9. Sex-limited and sex-influenced — which are not sex linkage

Sex-linked means the gene sits on a . That is a statement about location, and it is Ch 13. The two phenomena here involve ordinary autosomal genes whose map to takes sex as an argument.

Sex-limited: expressed in one sex only. Milk yield is an autosomal trait carried and transmitted by bulls who never express it. Familial male-limited precocious puberty is caused by activating variants in LHCGR, is autosomal , and produces a phenotype only in boys — it passes silently through generations of women. Sex-limitation is simply conditioned on sex, with one of the conditional penetrances at zero.

Sex-influenced: expressed in both sexes but with different dominance or a different threshold. In some sheep breeds the horned behaves as dominant in rams and in ewes, so heterozygotes are horned if male and polled if female — one , two phenotypes, and the difference is hormonal rather than genetic. Pattern baldness is the standard textbook example and the standard textbook example is wrong: androgenetic alopecia is polygenic, with a large contribution from the AR region. Use the sheep.

10. Environment, phenocopies, and conditional alleles

Generalise: the map takes environment as an argument too. A norm of reaction is as a function of environment for a fixed , and different genotypes have differently shaped functions — which is genotype-by-environment interaction, and is the general case rather than a complication.

Temperature-sensitive make this literal. The Himalayan rabbit and the Siamese cat carry a TYR allele encoding a tyrosinase that is inactive at core body temperature and active a few degrees below it. Pigment therefore appears only where the animal is cool: ears, muzzle, paws, tail. The demonstration is the classic one — shave a patch on a Himalayan rabbit's back, keep it cold while the hair regrows, and the new hair comes in black. Same genotype, same animal, different phenotype, decided by a runtime parameter.

Conditional alleles are also a tool: temperature-sensitive mutants let you study genes whose complete loss is lethal, by growing at permissive temperature and shifting (Ch 37).

Phenylketonuria carries the same lesson with the sign reversed. The genotype's effect depends entirely on dietary phenylalanine; restrict the diet from birth and the neurological phenotype largely does not appear. A fully genetic disease with an environmental treatment. "Genetic" does not mean "fixed".

Phenocopy: an environmentally produced phenotype indistinguishable from a genetic one. Thalidomide-induced limb reduction mimicked a rare inherited malformation closely enough to delay recognition of the cause; congenital rubella deafness mimics genetic deafness; nutritional rickets mimics hypophosphataemia. The consequence for Ch 15 is direct — an affected individual in a is not proof of a genotype.

Genetic background and modifier genes. The same allele on different backgrounds gives different phenotypes. In mice this is routine and infuriating: that are embryonic-lethal on one inbred strain are viable on another. In humans, CFTR F508del homozygotes vary enormously in lung disease severity, and association studies in cystic fibrosis cohorts have mapped modifier loci including SLC26A9, SLC9A3 and TGFB1. The best-documented case is sickle-cell disease, where severity is strongly modified by variants at BCL11A and HBS1L–MYB that keep fetal haemoglobin switched on — and BCL11A's erythroid is now the target of an approved gene-editing therapy (Ch 38). A modifier gene became a drug target.

— earlier onset and greater severity in successive generations — looked for decades like a violation of Mendelian logic, and turned out to have a physical cause: unstable . The repeat tract changes length during transmission, and length predicts onset. Myotonic dystrophy type 1 (CTG in the DMPK 3′UTR) expands most dramatically through the mother, producing the congenital form; Huntington disease (CAG in HTT) expands most through the father; fragile X (CGG in FMR1) converts to full only through the mother. Two cautions: anticipation is also generated artefactually by ascertainment, since you notice a severely affected child and then look backwards; and the molecular sizing is what turns an appearance into a finding. Ch 20 develops the mechanism, and D3, in the optional track, develops the disorder class — including why anticipation is real in some repeat diseases and absent in others.

11. What actually changed

PhenomenonWhat changed
intactintactmap became one-to-one; 1:2:1
intactintactboth products visible
Multiple intactintactmore than two alleles in the population
intactintactone class removed before you count
intactintactclasses merged by pathway topology
intactintactmap is stochastic
Sex-limited / -influencedintactintactmap takes sex as an input
Environmental modificationintactintactmap takes environment as an input

Not one row perturbs . The only phenomenon in Part 2 that genuinely breaks a Mendelian rule is , which breaks independent assortment because loci on the same are not sampled independently — and that is Ch 14.

Extend the last three rows to many loci at once, add environmental input and stochastic noise, and you have stopped doing transmission genetics and started doing quantitative genetics (Part 6). This chapter is the bridge, and the ratios are the last place the discrete framework is still visible.


Common misconceptions

What people thinkWhat's actually true
and are two names for the same thingIncomplete dominance gives one intermediate ; codominance gives both phenotypes at once. Pink flowers versus a red cell carrying both M and N antigens
A suppresses or overpowers the oneNeither allele acts on the other. Recessivity is usually the by-product of a saturating flux curve: half the enzyme gives nearly all the flux
Dominance is a property of an alleleIt is a property of (allele, phenotype, assay). HbS is recessive clinically, codominant on a gel, and dominant for malaria resistance
A recessive is recessive in every respectAy is dominant for coat colour and recessive for lethality, from a single deletion. Dominance is indexed by phenotype
is dominance between genesDominance is a relationship between alleles at one locus; epistasis is between loci. The words are not parallel
Classical epistasis and statistical epistasis are the sameMasking in a pathway and a non-zero interaction term in a regression are different objects. Either can occur without the other, and the statistical version is scale-dependent
Modified ratios show Mendel was wrongEvery one of them sums to sixteenths, because produced exactly the classes Mendel predicted. Only the rendering changed
An unaffected person with a dominant disease allele must be a non-carrier, or the test was wrongIncomplete . They carry the and can transmit it. Obligate carriers are how it is recognised in
Penetrance is a single number for a variantIt is age-dependent, background-dependent, and upward-biased when estimated from families ascertained through affected members
Sex-limited traits are sex-linkedSex-linked is about chromosomal location. Sex-limited and sex-influenced traits are usually autosomal; the phenotype depends on hormonal context
"Genetic" means unchangeablePKU is fully genetic and largely preventable by diet. Genotype fixes a norm of reaction, not an outcome
Failure to complement proves the are in the same geneNot if either is dominant, and not under non-allelic non-complementation between partners in a complex
is inheritance of acquired severityIt is repeat-tract instability during transmission — a physical change in DNA length, plus a substantial dose of

Worked example

Part A — inferring architecture from a ratio

Two true-breeding white-flowered lines are crossed. F1 is entirely purple. F1 selfed gives F2: 264 purple, 205 white.

Step 1 — read the F1. Two white parents produced purple offspring. The complement, so they are in different genes. Call them C and P; parents are CC pp and cc PP, F1 is Cc Pp.

Step 2 — count the classes. Total = 469. Two phenotypic classes only, so the map is a Boolean function of two inputs. Candidate ratios in sixteenths: 9:7, 12:4 (= 3:1), 15:1, 13:3.

Step 3 — test 9:7.

expected purple = 469 × 9/16 = 263.8
expected white  = 469 × 7/16 = 205.1

χ² = (264 − 263.8)²/263.8 + (205 − 205.1)²/205.1
   = 0.00015 + 0.00005
   = 0.0002        df = 1        p ≈ 0.99

Step 4 — reject the alternatives. Under 15:1, expected purple is 439.7 and χ² is enormous. Under 3:1, expected purple is 351.8, giving χ² ≈ 87.6. Only 9:7 survives.

Step 5 — state the biochemistry. 9:7 means the three non-C_P_ classes are phenotypically identical: an AND gate. Two sequential steps in the anthocyanin pathway, both required, and failure at either gives the same colourless end point.

Step 6 — make a prediction that could falsify it. Testcross the F1 to cc pp. Gametes from Cc Pp are CP : Cp : cP : cp at 1:1:1:1, and only CP gives a Cc Pp offspring. Predicted 1 purple : 3 white. A competing 15:1 model would predict 3 purple : 1 white — a decisive experiment, and the reason you do it rather than stopping at the χ².

Step 7 — note what a different answer would have meant. Had F2 come out 9 purple : 3 pink : 4 white, the topology would be sequential rather than AND: gene C makes a pink intermediate, gene P converts pink to purple, and losing C gives white regardless of P. Same two genes, one edge moved, a different ratio.

Part B — penetrance arithmetic

A has 70% and is rare enough that unrelated mates are non-carriers.

Expected proportion affected among children of an affected heterozygote × unaffected homozygote. Transmission and expression are independent:

P(affected child) = P(inherits allele) × P(penetrant)
                  = 0.50 × 0.70
                  = 0.35

Not 0.50. Fitting a with a 1:1 expectation would reject dominance for no good reason.

A clinically unaffected child of that couple — what is the probability they carry the allele? Bayes, with C = carries:

P(C) = 0.5                  P(unaffected | C)  = 0.30
P(¬C) = 0.5                 P(unaffected | ¬C) = 1.00

P(C | unaffected) = (0.5 × 0.30) / (0.5 × 0.30 + 0.5 × 1.00)
                  = 0.15 / 0.65
                  = 0.231

So an unaffected child still has a ~23% chance of carrying it, and a ~23% × 50% × 70% ≈ 8% chance of an affected child of their own. Incomplete penetrance does not remove risk from a pedigree; it hides it.

Connections