Contents
Problem set 02 — Mendelian genetics Problem set

Problem set 02 — Mendelian genetics

Covers Ch 09–13.

Attempt before revealing. Genetics is learned by calculating.

A standing note on method: are a teaching device that scales terribly. A trihybrid cross needs a 64-cell grid; a four-gene cross needs 256. Use the product rule on independent loci instead — it is faster, less error-prone, and generalises. If the product and sum rules are not yet second nature, they are §2 of S1. Several solutions below show both methods so you can see the saving.


1. Monohybrid, and what the ratio is actually telling you

In peas, purple flower (A) is to white (a). A true-breeding purple plant is crossed to a true-breeding white plant.

(a) Give the F1 and . (b) F1 × F1. Give F2 genotypic and phenotypic ratios. (c) You obtain 1,000 F2 plants. How many purple plants do you expect to be heterozygous? (d) You pick one purple F2 plant at random. What is the probability it is homozygous?

Solution

(a) AA × aa → all Aa, all purple.

(b) Aa × Aa:

Genotypic 1 AA : 2 Aa : 1 aa Phenotypic 3 purple : 1 white

(c) Of 1,000 F2, the expected genotype counts are 250 AA, 500 Aa, 250 aa.

Heterozygous purple = 500.

(d) This is a conditional probability, and it is the part people get wrong.

You have already observed that the plant is purple, which excludes aa. Among purple plants the genotypes are AA and Aa in ratio 1:2.

P(AA | purple) = 1/(1+2) = 1/3

The unconditional P(AA) is 1/4. Conditioning on the observed phenotype changes it to 1/3. This 1/3 recurs constantly in work — an unaffected sibling of someone with a condition has a 2/3 chance of being a carrier, by exactly this argument.


2. Dihybrid without a Punnett square

Seed shape (R round to r wrinkled) and colour (Y yellow dominant to y green) assort independently.

Cross RrYy × RrYy.

(a) P(round and yellow)? (b) P(wrinkled and green)? (c) P(round and green)? (d) P(RRYy) specifically? (e) What fraction of round yellow offspring are homozygous at both loci?

Solution

Treat the loci independently and multiply. From a monohybrid Aa × Aa: P(dominant ) = 3/4, P( phenotype) = 1/4.

(a) P(round) × P(yellow) = 3/4 × 3/4 = 9/16 (b) 1/4 × 1/4 = 1/16 (c) 3/4 × 1/4 = 3/16

Together with green round (3/16) these give the familiar 9:3:3:1, derived in three multiplications rather than a 16-cell grid.

(d) P(RR) × P(Yy) = 1/4 × 1/2 = 1/8

(e) Conditional again. Among round yellow offspring (9/16 of the total), the doubly homozygous RRYY occurs at 1/4 × 1/4 = 1/16.

P(RRYY | round yellow) = (1/16) ÷ (9/16) = 1/9


3. Trihybrid — where the method pays off ★

Cross AaBbCc × AaBbCc, all three loci independent.

(a) P(offspring showing all three )? (b) P( exactly AaBbCc)? (c) P( at exactly one of the three loci)? (d) How many cells would the need?

Solution

(a) (3/4)³ = 27/64 ≈ 0.422

(b) For each locus, P(heterozygous from Aa × Aa) = 1/2.

(1/2)³ = 1/8

(c) "Recessive at exactly one locus" means recessive phenotype at one, dominant at the other two — and the recessive locus could be any of the three, so this is a binomial term.

P(one specific locus recessive, others dominant) = (1/4) × (3/4) × (3/4) = 9/64

Three ways to choose which locus: 3 × 9/64 = 27/64 ≈ 0.422

Equivalently, using the binomial with n = 3, p(recessive) = 1/4:

P(exactly 1) = C(3,1) × (1/4)¹ × (3/4)² = 3 × 1/4 × 9/16 = 27/64 ✓

(d) Each parent produces 2³ = 8 gamete types, so the square is 8 × 8 = 64 cells.

You answered (a) through (c) with a handful of multiplications. This is why the product rule is the method actually used, and Punnett squares are scaffolding to be discarded.


4. Test cross

A purple-flowered plant of unknown is crossed to a white (aa) plant. The cross yields 47 purple and 53 white offspring.

(a) What was the unknown genotype? (b) What result would the alternative genotype have produced? (c) Why is a test cross preferred over selfing for this purpose?

Solution

(a) Roughly 1:1 purple:white. A test cross to a homozygous reveals the tester parent's gametes directly:

  • If AA: all gametes A, all offspring Aa, all purple
  • If Aa: gametes 1/2 A and 1/2 a, offspring 1/2 Aa purple and 1/2 aa white — 1:1

Observed 47:53 is consistent with 1:1, so the unknown parent is Aa.

(b) AA would have given 100 purple and 0 white. A single white offspring would have excluded it.

(c) Because the test cross gives a direct, high-contrast readout of gamete composition. Crossing to aa means the tester contributes only recessive , so the offspring reports the unknown parent's gamete genotype one-to-one.

Selfing an Aa plant gives 3 purple : 1 white, and selfing AA gives all purple — you can still distinguish them, but you must observe enough offspring to see a white one. The probability of missing it by chance in n offspring is (3/4)ⁿ, which is still 5.6% at n = 10. The test cross needs far fewer offspring for the same confidence because the expected difference is 50% versus 0%, not 25% versus 0%.


5. Chi-square done properly ★

A dihybrid F1 self-cross gives 556 offspring:

Observed
round yellow315
round green108
wrinkled yellow101
wrinkled green32

(a) State the null hypothesis and compute expected counts. (b) Compute χ². (c) State degrees of freedom and interpret against a 0.05 threshold. (d) What would a non-significant result license you to conclude?

Solution

(a) H₀: two independently assorting loci, each with complete dominance, giving 9:3:3:1.

Total 556, so expected = 556 × (9, 3, 3, 1)/16:

PhenotypeObservedExpected
round yellow315556 × 9/16 = 312.75
round green108556 × 3/16 = 104.25
wrinkled yellow101556 × 3/16 = 104.25
wrinkled green32556 × 1/16 = 34.75

(b) χ² = Σ (O − E)²/E

  • (315 − 312.75)²/312.75 = 5.0625/312.75 = 0.01619
  • (108 − 104.25)²/104.25 = 14.0625/104.25 = 0.13489
  • (101 − 104.25)²/104.25 = 10.5625/104.25 = 0.10132
  • (32 − 34.75)²/34.75 = 7.5625/34.75 = 0.21763

χ² = 0.470

(c) df = (number of categories) − 1 = 4 − 1 = 3. No parameters were estimated from the data — the 9:3:3:1 ratio comes from the genetic hypothesis, not from the observations — so there is no further subtraction. This is the most common df error in genetics: df is not n − 1 − (something) unless you actually fitted a parameter.

Critical value at α = 0.05, df = 3, is 7.815. Our χ² = 0.470 is far below it (p ≈ 0.93). Do not reject H₀. The data are consistent with 9:3:3:1.

(d) Very little, and this is the point of the problem.

Failure to reject is not confirmation. It means the data do not provide evidence against the model — which is a much weaker statement. Specifically:

  • The test has limited power. Real but modest departures ( with high recombination, mild differential viability) would not be detected at this sample size.
  • Many alternative models also predict approximately 9:3:3:1 and would equally fail to be rejected.
  • A very low χ² is itself worth noticing. This is precisely Fisher's critique of Mendel's data: across many experiments Mendel's results fit expectation better than chance should allow, suggesting unconscious selection or over-eager classification of ambiguous plants. A goodness-of-fit test has two tails in a sense — data can be suspiciously good as well as suspiciously bad.

6. Binomial family composition

Two carrier parents (Aa × Aa) for a condition have four children.

(a) P(exactly one affected)? (b) P(at least one affected)? (c) P(the first two are unaffected and the last two affected)? (d) They already have three unaffected children. What is the probability the fourth is affected?

Solution

Each child is independently affected with p = 1/4.

(a) C(4,1) × (1/4)¹ × (3/4)³ = 4 × 0.25 × 0.421875 = 0.4219 = 27/64

(b) P(at least one) = 1 − P(none) = 1 − (3/4)⁴ = 1 − 81/256 = 175/256 ≈ 0.684

(c) A specified order, so no binomial coefficient:

(3/4) × (3/4) × (1/4) × (1/4) = 9/256 ≈ 0.0352

Compare with (a): the probability of "exactly two affected in any order" would be C(4,2) × (1/4)² × (3/4)² = 6 × 9/256 = 54/256. Specifying the order costs a factor of 6.

(d) 1/4.

has no memory. The three unaffected children do not change the probabilities for the fourth. This is the gambler's fallacy in a lab coat, and it comes up constantly in genetic counselling — families frequently believe they have "used up" their risk.

Note the contrast with problem 1(d): there, conditioning on an observation about the individual in question changed the probability. Here the observations are about different individuals and are independent, so they change nothing.


7. Multiple alleles and codominance

The ABO blood group has three : I^A and I^B are codominant with each other, and both are to i.

(a) List the for each of the four . (b) A type AB parent and a type O parent have children. What phenotypes are possible? (c) A woman of type A and a man of type B have a type O child. Give both parental genotypes. (d) Could that same couple have an AB child? With what probability?

Solution

(a)

PhenotypeGenotype(s)
AI^A I^A or I^A i
BI^B I^B or I^B i
ABI^A I^B
Oii

(b) I^A I^B × ii. Gametes: I^A or I^B from one parent, i from the other.

Offspring: 1/2 I^A i (type A) and 1/2 I^B i (type B).

Notably, neither AB nor O is possible — the children cannot have either parent's phenotype. A clean demonstration that "resembling a parent" is not what inheritance guarantees.

(c) A type O child is ii, so it received an i from each parent. Both parents must therefore carry i:

Mother I^A i, father I^B i.

(d) Yes. Each parent transmits I^A or i (mother) and I^B or i (father), each with probability 1/2.

P(AB) = P(I^A from mother) × P(I^B from father) = 1/2 × 1/2 = 1/4

The full offspring distribution is 1/4 I^A I^B (AB), 1/4 I^A i (A), 1/4 I^B i (B), 1/4 ii (O) — all four blood groups from one cross.


8. Epistasis — derive the ratio, don't memorise it ★

In Labrador retrievers, coat colour involves two loci. Locus B determines pigment type: B_ black, bb brown. Locus E controls whether pigment is deposited in the hair at all: E_ permits deposition, ee blocks it, giving yellow regardless of the B .

Cross BbEe × BbEe.

(a) Predict the phenotypic ratio from the pathway logic, without a . (b) Name the type of . (c) A yellow dog is crossed to a brown dog and all puppies are black. Give all three genotypes.

Solution

(a) Reason through the pathway. The E locus acts first — it gates whether any pigment reaches the hair — so evaluate it first.

  • P(ee) = 1/4 → yellow, regardless of B. That is 4/16.
  • P(E_) = 3/4 → pigment is deposited, and now B determines which:

    • P(B_) = 3/4 → black. 3/4 × 3/4 = 9/16
    • P(bb) = 1/4 → brown. 3/4 × 1/4 = 3/16

9 black : 3 brown : 4 yellow

The 4 is not a mysterious number: it is 3 + 1, the bb ee and B_ ee classes collapsing together because when pigment cannot be deposited, its colour is unobservable. Every "modified Mendelian ratio" is a 9:3:3:1 with classes merged by a pathway relationship, and reading it off the pathway is far more reliable than memorising a table.

(b) epistasis — the homozygous recessive ee genotype masks the B locus. The E gene is epistatic to B; B is hypostatic.

(c) All puppies black means every puppy is B_E_, so each parent must supply what the other lacks.

  • The yellow parent is ee by definition. Since no brown puppies appeared, and the other parent is bb, the yellow parent must supply B to every puppy: BBee.
  • The brown parent is bb. It must supply E to every puppy, so it is bbEE.
  • All puppies: BbEe — black.

A useful check: this is exactly the classic complementation pattern. Two differently-coloured parents producing uniformly black offspring tells you their defects are at different loci.


Where you went wrong

Error patternWhat to re-read
Used unconditional probability when the was already observedProblems 1(d), 2(e) — Ch 12
Applied a binomial coefficient to a specified birth orderProblem 6(c)
Thought previous children change the next child's riskProblem 6(d)
Got df wrong in the chi-squareProblem 5(c)
Read "not significant" as "hypothesis confirmed"Problem 5(d)
Memorised 9:3:4 instead of deriving it from the pathwayProblem 8(a) — Ch 11
Reached for a on three or more lociProblem 3