Contents
Problem set 05 — Mutation and chromosomes Problem set

Problem set 05 — Mutation and chromosomes

Covers Ch 16–20.

Attempt before revealing. Nearly every error in this set is a units error or a denominator error, and reading the solution first hides both from you.

Problems are roughly in order of difficulty; ★ marks the two hardest. Parts labelled Trap carry a misconception named explicitly in the solution.


1. Classify these five changes

Part of exon 2 of a gene, in frame, with the start of 2 in lower case. Exon 2 is 88 bp long and ends exactly at a boundary. Exon 3 begins GATCCTAAGCT... and is an internal exon — several more exons follow it.

pos:  1 2 3   4 5 6   7 8 9   10 11 12   13 14 15
      C A T   G G C   G A A   T  G  T    C  A  G  | g t a a g t ...
      His     Gly     Glu     Cys        Gln

For each change give (i) transition or transversion where the term applies, (ii) molecular class, (iii) consequence.

(a) pos 6, C→T (b) pos 7, G→T — Trap (c) pos 5, G→A (d) deletion of pos 10 (e) pos 15, G→A — Trap

Solution

Note the only CpG in the window, at positions 6–7 (GGCGAA). Three of the five changes touch it.

(a) C→T, →pyrimidine: transition. GGC → GGT, both Gly: synonymous, and here genuinely silent. It is the commonest de novo substitution in the genome — position 6 is a CpG cytosine, usually methylated, and deamination of 5-methylcytosine yields thymine directly — and it destroys the CpG, which is why observed CpG frequency (~1%) sits fourfold below expectation.

(b) Trap. G→T, →pyrimidine: transversion. GAA → TAA = STOP: nonsense, and — given the exon structure stated — far enough upstream of the final exon–exon junction to trigger : loss of transcript, not a truncated protein.

Do that step properly rather than by reflex, because the window puts the stop codon suspiciously close to a junction. Position 15 is the last base of an 88 bp exon 2, so the window is exon-2 positions 74–88 and the new TAA occupies positions 80–82 — only 88 − 82 = 6 nt from the exon 2/3 junction. The 50–55 nt rule is not measured against the nearest junction, though. The pioneer round of strips every exon-junction complex the passes through, so what decides the fate of the transcript is whether a complex is still deposited beyond the stop — i.e. whether the last junction lies more than 50–55 nt downstream of it. Exon 3 is internal, so the last junction sits far downstream and this PTC is degraded. Had exon 3 been the terminal exon, those same 6 nt would put the stop inside the last-junction window and the transcript would escape NMD — a truncated protein, the opposite prediction from the identical base change.

The trap: this is the G of the same CpG, so it looks like "the CpG on the other strand". It is not. On the antisense strand that G is a C, and its deamination product reads as G→A on the sense strand — a transition. G→T is not a deamination product at all. G:C→T:A comes from oxidation of guanine to 8-oxoguanine, which pairs with adenine — also the oxidised-library sequencing artefact and the signature of MUTYH loss.

(c) G→A, purine→purine: transition. GGC → GAC: Gly → Asp. , and non-conservative — glycine has no side chain and is the residue of choice at tight backbone turns.

(d) Deleting the T at position 10, spliced to exon 3:

before:  CAT GGC GAA TGT CAG | GAT CCT AAG CT...   His Gly Glu Cys Gln Asp Pro Lys
after:   CAT GGC GAA GTC AGG ATC CTA AGC T...      His Gly Glu Val Arg Ile Leu Ser

Frameshift. Everything from codon 4 on differs and the original stop is out of frame.

(e) Trap. G→A: transition. CAG → CAA, both Gln — synonymous at the protein level, and therefore exactly what a coding-only filter discards. It is a splice-site mutation. Position 15 is the last exonic base, position −1 of the 5′ donor site (consensus (C/A)AG | GURAGU), where the exon's terminal G is the most conserved exonic base and U1 snRNA pairs across the junction. Expect exon skipping, intron retention, or a cryptic donor (Ch 06).

Now use the exon length: skipping exon 2 removes 88 nt, and 88 / 3 = 29 remainder 1 — not a multiple of three, so skipping frameshifts everything downstream, the same consequence class as (d). Had exon 2 been 87 bp, skipping would have deleted 29 residues in frame, usually far milder.


2. Why transitions win when transversions outnumber them 2:1

(a) What Ti/Tv would uniform substitution give? Compute the enrichment implied by the observed genome-wide ~2.0–2.1. (b) Removing CpG C→T from a de novo spectrum drops Ti/Tv from ~2.1 to ~1.55. Derive what fraction of all CpG C→T must be. (c) Exome Ti/Tv is higher still, ~3.0–3.3. Give the reason in one sentence.

Solution

(a) Each of four bases can change to three others: 12 substitutions, four transitions (A↔G, C↔T), eight transversions.

null Ti/Tv = 4/8 = 0.5        2.1 / 0.5 = 4.2

Transitions run fourfold above chance. The 2:1 combinatorial handicap turns an observed 2:1 ratio into a fourfold excess; confusing ratio with enrichment is the standard error here.

(b) Let f = fraction of all mutations that are CpG C→T, all transitions. The remaining (1 − f) has Ti/Tv = 1.55, so within it transitions are 1.55/2.55 = 0.6078 and transversions 1/2.55 = 0.3922. Set the total ratio to 2.1:

[ f + 0.6078(1-f) ] / [ 0.3922(1-f) ] = 2.1

  0.3922f + 0.6078 = 0.8235 - 0.8235f
  1.2157f = 0.2157
        f = 0.177

About 17.7%, inside the 15–20% quoted from de novo studies — a real consistency check between two independently reported numbers. Note the sensitivity: with residual 1.6 and observed 2.0 the same algebra gives f = 0.133, so treat it as "roughly a sixth".

What f = 0.18 does not say: strip CpG C→T out entirely and Ti/Tv is still ~1.55, threefold above the null. CpG deamination carries about a third of the excess; the rest is the geometric ease of transition mispairing — a opposite a still fits the helix, while purine:purine and pyrimidine:pyrimidine mismatches distort it.

(c) , which acts on coding sequence and not on most of the genome. Classify third-position changes across the 61 sense — one transition and two transversions each:

transitions   synonymous: 58 / 61  = 95.1%
transversions synonymous: 68 / 122 = 55.7%

95% of third-position transitions are invisible to protein-level selection, against 56% of transversions — the gap comes from the twofold-degenerate boxes, where the synonymous partner is always a transition. So Ti/Tv needs different thresholds per capture region: a whole-genome callset at 2.0 is healthy, an exome callset at 2.0 is alarming.


3. How many mutations did you invent?

Pinned rate 1.1–1.3 × 10⁻⁸ per bp per generation; genome 6.2 × 10⁹ bp.

(a) Expected de novo point in a child. (b) Trap. Why 6.2 × 10⁹ and not 3.1 × 10⁹? (c) Trio studies report ~60–70. What callable fraction reconciles that? (d) Trap. Replication fidelity is ~10⁻¹⁰ per base. Why can you not use it here?

Solution

(a)

1.1e-8 x 6.2e9 = 68.2        1.3e-8 x 6.2e9 = 80.6

Roughly 68–81 new point mutations neither parent carried.

(b) Trap. Every base pair in the child is a transmission event and there are two copies of each: a mutation on either counts, and both are sequenced. Halving to 3.1 × 10⁹ gives 37, and comes from thinking of "the genome" as the 3.1 Gb reference rather than the 6.2 Gb the child was handed.

(c) Short reads call variants only in the callable genome — not , , or long tandem repeats. With ~15% unreadable:

0.85 x 80.6 = 68.5        0.85 x 68.2 = 58.0

which brackets the reported 60–70. Inverting instead, to reproduce 65 from the midpoint:

midpoint = (68.2 + 80.6)/2 = 74.4        65 / 74.4 = 0.874  ->  ~13% not callable

The naive product is the true number; the reported number is the true number minus what the instrument cannot see. This is also why the long-read estimate (1.30 × 10⁻⁸) sits at the top of the range: long reads recover the repetitive fraction, which is also the fastest-mutating part.

(d) Trap. Because 10⁻¹⁰ is per base per replication and 1.2 × 10⁻⁸ is per base per generation. They differ ~100× and are not interchangeable.

Used directly, 10⁻¹⁰ × 6.2 × 10⁹ = 0.62 mutations per cell division — true, and not what was asked. A generation is not one replication. Summing ~300 germline divisions:

300 x 1e-10 = 3e-8 per bp per generation

against a measured 1.2 × 10⁻⁸ — an overshoot of ~2.5×, the expected direction, because 10⁻¹⁰ is an order-of-magnitude bound rather than a constant. Pushing the other way, much germline mutation never was a replication error: unrepaired deamination and depurination, which no polymerase fidelity would prevent. Agreement within a factor of two to three is all this cross-check can claim.


4. Paternal age

Pinned: ~1.3–1.5 extra per year of paternal age, ~80% paternal in origin. Anchor the total at 70 at paternal age 30.

(a) Expected count for fathers aged 20 and 40. (b) Paternal age has sd 6 years in a cohort. Predict the variance of the de novo count across trios and compare with Poisson.

Solution

(a) Slope 1.4/year, anchored at 70 at age 30:

age 20:  70 - 10 x 1.4 = 56
age 40:  70 + 10 x 1.4 = 84        difference 28,  ratio 1.5

Across the pinned range the difference spans 26 (at 1.3/yr) to 30 (at 1.5/yr). A 40-year-old father transmits ~28 more de novo mutations than a 20-year-old — 50% more than that 20-year-old's 56.

Name that denominator, because Ch 16 §7 quotes the same 28 as a ~40% increase: 28/56 = 0.50 against the younger father's own count, 28/70 = 0.40 against the population baseline of 70 at age 30. Same difference, two denominators, two percentages, neither wrong — and both uninterpretable unstated.

Split by parent, this also shows the 80% figure is not a constant. At age 30, 0.80 × 70 = 56 paternal and 14 maternal; holding maternal at 14 and putting the whole slope on the father gives 42/56 = 75% paternal at age 20 and 70/84 = 83% at age 40.

Sanity-check against divisions. Spermatogonia divide ~23 times a year, so a pure replication-error account at 10⁻¹⁰ per base over a haploid 3.1 × 10⁹ bp genome predicts 23 × 1e-10 × 3.1e9 = 7.13 mutations/yr against an observed 1.4 — an overshoot of ~5×. The age effect is not purely replicative.

(b) Condition on paternal age:

Var(N) = E[Var(N|age)] + Var(E[N|age])
       = 70 + (1.4 x 6)^2 = 70 + 70.56 = 140.6      sd = 11.9

Index of = 140.6/70 = 2.0: the count is twice as dispersed as Poisson, and essentially all the excess is one covariate. Same second-moment reasoning as the Luria–Delbrück — the mean tells you little, the variance identifies the process.

The trap: none of this transfers to the mother. Maternal age is the risk factor for , via loaded in fetal life and decaying over decades of prophase-I arrest (Ch 20 §6) — errors, not point mutations.


5. Lesion → pathway → disease

(a) For each lesion name the pathway and the disease caused by losing it: (1) UV cyclobutane dimer; (2) adenine mis-inserted opposite 8-oxoguanine; (3) single-base mismatch and small loops at a (A)₂₅ tract; (4) O⁶-methylguanine from temozolomide; (5) two-ended double-strand break in G1; (6) interstrand crosslink from mitomycin C. (b) An identical break occurs in G1 and in S . What single step decides the pathway, and what gates it? (c) Trap. XP and Cockayne syndrome are both NER defects. XP carries >10,000-fold skin cancer risk; CS carries essentially none. Reconcile. (d) Cytosine deamination produces 100–500 uracils per cell per day. How efficient must uracil excision be?

Solution

(a)

#PathwayDisease of loss
1 — recognises helix distortion, not the chemistryXeroderma pigmentosum; losing only the -coupled branch gives Cockayne syndrome
2, glycosylase MUTYHMUTYH-associated polyposis, with excess G:C→T:A
3 (MutSα/MutSβ, MutLα)Lynch syndrome; biallelic loss gives CMMRD. Readout is microsatellite instability
4Direct reversal by MGMT, a suicide enzyme — stoichiometric, not catalyticMGMT methylation predicts temozolomide response in glioblastoma
5Non-homologous end joiningRadiosensitive SCID (LIG4, PRKDC, NHEJ1, DCLRE1C) — V(D)J recombination is , so the repertoire fails with the repair
6Interstrand crosslink / Fanconi pathway, feeding into HRFanconi anaemia

(b) The committing step is 5′→3′ end resection. Once ends are single-stranded Ku cannot bind and NHEJ is off the table; RPA coats the ssDNA and BRCA2 loads RAD51. Resection is gated by CDK activity — low in G1, high from S on — which phosphorylates CtIP and EXO1; on top, 53BP1–RIF1–shieldin blocks resection and BRCA1–BARD1 licenses it. The elegance: the licensing signal is the availability of the template, since sisters exist only after replication.

(c) Trap. Not "XP loses one branch, CS the other" — that fits only XP-C and XP-E. XP-A, XP-B, XP-D, XP-F and XP-G lose core incision factors shared by both branches, so five of seven groups lose transcription-coupled NER too.

What separates them is the fate of the cell after repair fails. In XP lesions persist in dividing cells, get bypassed by error-prone translesion polymerases, and become — cancer. In CS the lesions that matter stall II, a potent apoptotic signal, so cells die or arrest: tissue attrition and ageing instead of tumours. A repair deficiency causes cancer only in cells that survive it.

(d) Take 300/day over a 70-year lineage, allowing this one reaction at most ~100 lifetime substitutions:

300 x 365 x 70 = 7,665,000 deamination events
100 / 7.665e6  = 1.30e-5     <- escape fraction
1 - 1.30e-5    = 0.9999870   <- 99.9987% success

Uracil excision must succeed better than 99.998% of the time, continuously, for seventy years — which it does, precisely because uracil is chemically foreign. For 5-methylcytosine, whose product is thymine, no unambiguous recognition is available; that loss of detectability is the entire CpG hotspot, whose size problem 2(b) measured.


6. ★ Why a PARP inhibitor kills one cell and not its neighbour

A patient with a pathogenic BRCA2 variant is treated with olaparib.

(a) Why is the tumour HR-deficient while the patient's normal cells are not — and when did that asymmetry arise? (b) Give the mechanism connecting PARP inhibition to death of an HR-deficient cell. (c) Model the window. If PARP inhibition generates N one-ended double-strand breaks per S , resolved with probability 0.999 by an HR-proficient cell and 0.5 by an HR-deficient one, compute survival for N = 10 and N = 50. (d) Trap. The tumour relapses with biallelic loss of TP53BP1. Explain — carefully.

Solution

(a) The patient is germline heterozygous: one working suffices, so every normal cell does HR normally. The tumour arose after somatic loss of the second allele, so it alone is HR-deficient. The asymmetry exists before any drug is given — the drug does not create selectivity, it exploits a difference tumour evolution already manufactured. Hence a germline companion diagnostic on blood.

(b) : losing either gene alone is survivable, losing both is not.

BRCA2-/-              viable   (true by construction — the tumour exists)
PARP1 inhibited       viable   (normal cells tolerate it)
BRCA2-/- + PARP1 off  dead

PARP1 detects single-strand breaks and nucleates their repair. Inhibit it and SSBs persist; a hitting one collapses into a one-ended double-strand break. One-ended breaks cannot go through — there is no second end to ligate to, and joining it to any other free end makes a — so the only correct route is HR off the sister chromatid. The HR-deficient cell has none: it dies, or repairs by NHEJ/MMEJ into fusions and dies at mitosis anyway.

The drug's target is intact in the tumour. PARP1 is not mutated; the relationship is between two different genes, which is what "synthetic" means. (Killing actually tracks PARP trapping on DNA better than catalytic inhibition — talazoparib and veliparib inhibit comparably yet differ ~100-fold in potency.)

(c) Survival across one S phase is the per-break success probability to the N:

N = 10:  proficient 0.999^10 = 0.9900     deficient 0.5^10 = 9.77e-4    ratio 1.01e3
N = 50:  proficient 0.999^50 = 0.9512     deficient 0.5^50 = 8.88e-16   ratio 1.07e15

A per-break difference of a factor of two becomes a selectivity of 10³ at ten breaks and 10¹⁵ at fifty, because survival compounds multiplicatively over independent breaks — the fidelity-as-a-product argument of Ch 04 run in reverse. Selectivity therefore depends steeply on dose, which sets N, and collapses as the deficient cell's per-break success climbs back toward 1.

(d) Trap. BRCA1's essential HR function is to antagonise 53BP1–RIF1–shieldin and license resection, so deleting 53BP1 lets resection proceed without BRCA1, RAD51 loads, and HR is restored — a well-documented resistance route.

But this patient is BRCA2-mutant, and BRCA2 acts downstream of resection: it loads RAD51 onto RPA-coated ssDNA. Deleting 53BP1 in a BRCA2-null cell gives beautifully resected ends that still cannot become a RAD51 filament. HR stays dead, the cell stays sensitive. So the finding as stated does not explain this relapse — look instead for a BRCA2 reversion , restored fork protection, or drug efflux. The trap is treating "BRCA1/2" as one gene; the asymmetry of 53BP1 rescue is the cleanest evidence they act at different steps.


7. Reading nondisjunction off a genotype

A pericentromeric STR on 21, 200 kb from the . Read depths in four unrelated 21 probands:

FamilyMotherFatherChild (fraction of reads)
I28/3032/3228 (33%), 30 (33%), 32 (33%)
II28/3032/3228 (67%), 32 (33%)
III28/2832/3428 (33%), 32 (33%), 34 (33%)
IV28/3032/3228/30/32 in 30% of cells; 28/32 in the rest

(a) Assign an origin to each. (b) Trap. Crossovers between marker and centromere are rare but possible. With P(signature-flipping crossover) = 0.05 and P(MI) = 0.75 among maternal errors: given an observed centromere-homozygous result, what is P(true origin was MI)? Given heterozygous, P(MII)?

Solution

The logic in one line: MI sends one chromatid of each homologue into the gamete; MII sends two sister chromatids of one homologue. So count the distinct alleles contributed by the erring parent.

(a)

  • I — both maternal alleles at equal dose: centromere-heterozygous → maternal MI.
  • II — the maternal contribution is 28 at double dose. The 67:33 read ratio is the assay; without dosage this looks like an ordinary heterozygote. Centromere-homozygous → maternal MII.
  • III — the mother is uninformative (28/28) but the father contributed both 32 and 34: paternal MI. A marker reports only on the parent heterozygous at it.
  • IV — two cell lines: post-zygotic mitotic nondisjunction, i.e. mosaic trisomy 21.

For calibration: ~90% of trisomy 21 is maternal and ~70–80% of those are MI, so of 100 cases expect 67.5 maternal MI, 22.5 maternal MII, 10 other — call it ~68 / ~22 / ~10, and note that rounding both halves up gives 101 cases out of 100.

(b) Trap. One crossover between centromere and marker swaps the signatures. With P(MI) = 0.75, P(MII) = 0.25, c = 0.05:

P(homozygous)   = 0.75 x 0.05 + 0.25 x 0.95 = 0.0375 + 0.2375 = 0.2750
P(MI | homozygous)    = 0.0375 / 0.2750 = 0.136

P(heterozygous) = 0.75 x 0.95 + 0.25 x 0.05 = 0.7125 + 0.0125 = 0.7250
P(MII | heterozygous) = 0.0125 / 0.7250 = 0.017

A single-marker "MII" call is wrong ~14% of the time; a single-marker "MI" call is wrong ~2%. The asymmetry is pure base rate — MI is three times more common, so its rare leakage into the small homozygous class rivals the genuine MII cases, while MII's leakage into the large heterozygous class is swamped. Nothing about the assay is asymmetric; the prior is. Hence real studies a panel. And even a correct centromere-homozygous call does not cleanly mean MII — much of that class is attributed to premature separation of sister chromatids during MI, so report what you measured and treat "MI/MII" as a model laid on top.


8. ★ Inversion mechanics

Two carriers, each heterozygous for an . ● is the .

Carrier P (paracentric)          Carrier Q (pericentric)
normal    1 2 ● 3 4 5 6 7 8      normal    1 2 3 ● 4 5 6 7
inverted  1 2 ● 6 5 4 3 7 8      inverted  1 2 5 4 ● 3 6 7

In each, one crossover occurs inside the inversion loop, between markers 4 and 5.

(a) Enumerate the four chromatids for carrier P. (b) Enumerate the four chromatids for carrier Q. (c) Why does one carrier's risk present as reduced fertility and the other's as an abnormal liveborn? Which inversion size is more dangerous? (d) Trap. Why is measured recombination across an inversion near zero?

Solution

Aligning an inverted segment with its normal partner forces a loop: markers pair up 1,2,3,4,5,… along the axis, but on the inverted that axis runs backwards along the physical molecule. Everything follows from tracing the four arms leaving the exchange point, which joins the axis-decreasing arm of one chromatid to the axis-increasing arm of the other.

(a) Paracentric. Split each chromosome physically at the crossover — between 4 and 5 on the normal, between 5 and 4 on the inverted, which reads 1 2 ● 6 5 | 4 3 7 8:

recombinant 1 = (1 2 ● 3 4) + (5 6 ● 2 1) = 1-2-●-3-4-5-6-●-2-1
                TWO centromeres -> DICENTRIC     dup 1,2 ; no 7,8

recombinant 2 = (8 7 3 4)   + (5 6 7 8)   = 8-7-3-4-5-6-7-8
                NO centromere  -> ACENTRIC       dup 7,8 ; no 1,2

The other two chromatids were uninvolved and are parental — one normal, one inverted, both balanced. The dicentric bridges and breaks; the acentric is never captured by the spindle and is lost. The duplication/deficiency is present, simply packaged in structures that destroy themselves.

(b) Pericentric. Same procedure; the inverted chromosome reads 1 2 5 | 4 ● 3 6 7:

recombinant 1 = (1 2 3 ● 4) + (5 2 1)   = 1-2-3-●-4-5-2-1    ONE centromere, dup 1,2 ; del 6,7
recombinant 2 = (7 6 3 ● 4) + (5 6 7)   = 7-6-3-●-4-5-6-7    ONE centromere, dup 6,7 ; del 1,2

Two parental chromatids again, but each recombinant is now monocentric and perfectly transmissible, carrying a duplication of the material distal to one breakpoint and a deficiency of that distal to the other.

(c) The centromere's position relative to the loop is the whole difference. Paracentric products self-destruct before fertilisation, so the carrier shows near-normal reproductive outcomes, at most slightly reduced fertility. Pericentric products are viable chromosomes that fertilise, so the carrier presents with miscarriage or an unbalanced liveborn.

Note which material is imbalanced: always the flanking regions outside the inversion. Hence the counterintuitive risk rule — a larger inverted segment captures more crossovers and leaves smaller flanking segments, so the imbalances are smaller and more survivable. Risk is negligible below ~30% of chromosome length and highest above 50%: the dangerous inversion is the big one.

(d) Trap. It does not mean crossovers fail to happen. They happen. What is suppressed is the recovery of the products — dicentrics and acentrics are destroyed, unbalanced conceptuses are lost. Either way the recombinants never reach the next generation, so the observed recombination fraction collapses toward zero and the inverted segment is inherited as one non-recombining block with persistently high (Ch 29). Keep the two risks apart, though: a balanced carrier has normal gene dosage and is typically healthy. "Balanced" describes the carrier's genome, not their gametes.


Where you went wrong

Error patternWhat to re-read
Treated a synonymous change as consequence-freeProblem 1(e) — the exon's last base is part of the donor site
Called G→T at a CpG "the CpG "Problem 1(b) — deamination gives G→A
Judged NMD against the nearest junction rather than the last oneProblem 1(b) — the stop sits 6 nt from the exon 2/3 junction
Confused the raw with enrichment over chanceProblem 2(a) — the null is 0.5
Halved the genome when counting Problem 3(b) — both transmitted copies count
Used per-replication fidelity for a per-generation questionProblem 3(d), and verified facts
Quoted a percentage change without naming its denominatorProblem 4(a) — 28 mutations is +50% or +40% depending on the base
Blamed maternal age for a child's extra point mutationsProblem 4(b) — maternal age drives
Said XP and Cockayne differ by which NER branch is lostProblem 5(c) — five of seven XP groups lose both
Assumed 53BP1 loss rescues any BRCA-mutant tumourProblem 6(d) — BRCA1 only; BRCA2 acts downstream
Read one pericentromeric marker as a definitive MI/MII callProblem 7(b) — ~14% of "MII" calls are MI
Said prevent Problem 8(d) — they prevent recovery of the products