Problem set 01 — Molecular foundations
Covers Ch 02–08.
Attempt each problem before opening the solution. Reading a worked solution produces a strong and completely false sense of understanding. If you are stuck, re-read the chapter's worked example rather than revealing the answer.
Problems are roughly in order of difficulty. ★ marks ones worth returning to.
1. Strands and direction
You are given the sequence 5'-ATGCGTACGGATTAGCC-3'.
(a) Write its complementary strand, correctly labelled with 5′ and 3′ ends. (b) Write the reverse complement in the conventional 5′→3′ orientation. (c) Compute the GC content. (d) A colleague sends you this sequence in a file with no strand and asks which strand of the it came from. What do you tell them?
Solution
(a) Pair each base and note the strands are antiparallel:
5'- A T G C G T A C G G A T T A G C C -3'
3'- T A C G C A T G C C T A A T C G G -5'
(b) Reverse the complement so it reads 5′→3′: 5'-GGCTAATCCGTACGCAT-3'
A useful check: the reverse complement of the reverse complement returns the original.
(c) Count G and C: positions 3, 4, 5, 8, 9, 10, 15, 16, 17 — nine of seventeen.
GC = 9/17 = 52.9%
(d) You cannot tell, and neither can they. A double-stranded DNA molecule has two strands, and either can be written down; "the sequence" is meaningless without a reference coordinate and a strand. This is exactly why FASTA of a genomic region is ambiguous unless you know the assembly and orientation, and why annotation formats such as GFF carry an explicit strand column (Ch 41).
The deeper point: strandedness is a property of the annotation, not of the molecule.
2. Melting behaviour
Two DNA duplexes of identical length, 60 bp:
- Duplex A: 30% GC
- Duplex B: 70% GC
(a) Which has the higher , and why? (b) Give two distinct physical reasons, not one. (c) You need to design primers for a GC-rich region. What problem does this create and how would you address it?
Solution
(a) Duplex B, the GC-rich one.
(b) Two reasons, and most students give only the first:
- Hydrogen bonding. G–C pairs form three hydrogen bonds, A–T only two, so more energy is needed to separate the strands.
- . GC-containing steps have more favourable stacking interactions than AT-containing ones. This contributes more to duplex stability than the hydrogen-bond difference does — a point Ch 02 makes and which is routinely omitted.
The common answer "because G–C has three hydrogen bonds" is correct but incomplete, and it propagates the misconception that pairing dominates stability. It doesn't; stacking does.
(c) GC-rich templates resist denaturation, so the strands do not separate cleanly at standard cycling temperatures, and they also form stable secondary structures (hairpins, G-quadruplexes) that block polymerase progression. Practical responses: raise the denaturation temperature, add a co-solvent such as DMSO or betaine to destabilise secondary structure, use a polymerase and buffer formulated for GC-rich templates, and design primers to avoid self-complementarity. This is also why GC-extreme regions are systematically under-covered in sequencing libraries that involve amplification (Ch 40).
3. Replication arithmetic ★
The human genome is approximately 3.1 × 10⁹ bp per haploid set. Human move at roughly 2 kb per minute, and S lasts about 8 hours.
(a) If replication proceeded from a single bidirectional origin, how long would copying one copy of the genome take? (b) Given that S phase is ~8 hours, estimate the minimum number of origins required. (c) Explain why the actual number of origins used is far larger than your estimate.
Solution
(a) A single bidirectional origin has two forks moving in opposite directions, so together they cover 2 × 2 kb = 4 kb per minute.
Time = 3.1 × 10⁹ bp ÷ 4 × 10³ bp/min = 7.75 × 10⁵ minutes
= 7.75 × 10⁵ ÷ 60 ≈ 12,900 hours ≈ 538 days
Roughly a year and a half to copy one genome. This is the calculation that makes the necessity of multiple origins vivid.
(b) With 8 hours = 480 minutes available, each bidirectional origin can cover:
480 min × 4 kb/min = 1,920 kb ≈ 1.92 Mb
Minimum origins = 3.1 × 10⁹ ÷ 1.92 × 10⁶ ≈ 1,615 origins
(c) Several reasons, and they compound:
- Origins are not evenly spaced or uniformly efficient. Licensing is stochastic; a given licensed origin fires in only a fraction of cell cycles. To guarantee everywhere, the cell must license far more origins than the minimum.
- Forks stall and collapse. Replication runs into damage, complexes, and hard-to-replicate structures. Dormant origins provide backup — if a fork dies, a nearby origin can fire and rescue the region. A system with exactly the minimum number of origins would fail whenever any fork failed.
- Timing is programmed. Replication is not uniform in time; euchromatin replicates early and late, which requires more origins than a uniform schedule would.
Estimates for licensed origins in a human cell run to tens of thousands, with a smaller subset firing in any given cycle. The redundancy is the point: this is a system engineered for robustness against fork failure, not for minimal resource use.
4. Okazaki fragments
Consider a single replication bubble with two bidirectional forks that together replicate 120 kb of DNA. Human are roughly 150 nucleotides.
(a) How much of the 120 kb is synthesised discontinuously? (b) Approximately how many Okazaki fragments are produced? (c) How many ligation events are needed?
Solution
(a) At each fork one strand is leading and one is lagging. Both parental strands are copied, so across the whole bubble exactly half the newly synthesised DNA is made discontinuously: 60 kb.
The symmetry is worth pausing on. It is a direct consequence of antiparallel strands plus a polymerase that only works 5′→3′; there is no way to arrange it otherwise.
(b) 60,000 nt ÷ 150 nt per fragment = 400 Okazaki fragments
(c) Each fragment must be joined to the one behind it. For 400 fragments in a bubble, the number of joins is 400 minus the number of fragment runs that terminate at a bubble end. With two (one per fork), you have two runs of ~200 fragments each, needing ~199 joins apiece, so approximately 398 ligation events — and in practice the final fragments are joined to the neighbouring replicon rather than left free.
The order-of-magnitude answer, ~400, is the one that matters. The point of the calculation is that lagging-strand synthesis imposes an enormous number of priming, primer-removal, gap-fill and ligation events, every one of which is an opportunity for error — which is why the lagging strand carries a slightly different from the leading strand.
5. Fidelity as a product ★
DNA replication fidelity is achieved by three mechanisms acting in series:
| Mechanism | Error rate after this step |
|---|---|
| Polymerase base selection | ~1 in 10⁵ |
| 3′→5′ exonuclease | ~1 in 10⁷ |
| ~1 in 10¹⁰ |
(a) By what factor does each successive mechanism improve fidelity? (b) How many replication errors would you expect per genome (6.2 × 10⁹ bp) copied? (c) A tumour has lost mismatch repair function. Estimate the resulting error rate per genome replication, and name the diagnostic signature this produces. (d) ★ The is ~1.2 × 10⁻⁸ per bp per generation. Why is it not a contradiction that this is a hundred times worse than the per-replication fidelity above?
Solution
(a) Roughly 100-fold from proofreading and 1,000-fold from mismatch repair:
- Base selection alone: 10⁻⁵
- Proofreading: 10⁻⁵ → 10⁻⁷, a 10²-fold improvement
- Mismatch repair: 10⁻⁷ → 10⁻¹⁰, a further 10³-fold
No individual mechanism is remotely accurate. The final fidelity is the product of three mediocre filters — the general shape of every fidelity number in biology (Ch 01 §5).
(b) 6.2 × 10⁹ bp × 10⁻¹⁰ errors/bp = ~0.6 errors per diploid genome replication
Fewer than one substitution per complete duplication of the genome. Worth pausing on: copying 6.2 billion characters and expecting to get all but roughly half of one right.
(c) Losing mismatch repair removes the last 1,000-fold filter, so the error rate returns to approximately 10⁻⁷ per bp:
6.2 × 10⁹ × 10⁻⁷ = ~620 errors per genome replication
A thousandfold increase, compounding at every division — a hypermutator .
(d) Because they are different units measuring different things, and this is the most common unit error in the whole subject.
- 10⁻¹⁰ is per base per replication — one cell division.
- 1.2 × 10⁻⁸ is per base per generation — the whole path from zygote to gamete.
A generation is not one replication. The male germline undergoes a few hundred divisions between zygote and sperm, so replication errors alone accumulate to roughly 10⁻¹⁰ × 300 ≈ 3 × 10⁻⁸ per generation — the same order as the measured rate.
Note the direction of the disagreement: the estimate overshoots the measurement by a factor of two or three. That is the expected failure mode, because 10⁻¹⁰ is an order-of-magnitude bound rather than a measured constant. And pushing the other way, a substantial share of germline is not replicative at all — it comes from unrepaired chemical damage such as deamination and depurination (Ch 16), which no amount of polymerase fidelity would prevent.
Agreement within a small factor is all this calculation is entitled to claim. Quoting the two numbers as though they were comparable is not.
The diagnostic signature is microsatellite instability. Mismatch repair is what corrects the strand-slippage errors that polymerase makes at short tandem repeats, so without it, microsatellite tract lengths become unstable and vary between cells. MSI is used clinically to identify Lynch syndrome and, importantly, as a biomarker predicting response to immune checkpoint inhibitors — because a hypermutated tumour generates many neoantigens (Ch 56).
6. Transcription and strand conventions
A gene is annotated on the reverse strand. The forward-strand genomic sequence across the first part of its coding region reads:
5'-...GGCATCGATTACGCAT...-3' (forward strand, GRCh38)
(a) Which strand serves as the template for ? (b) Write the mRNA sequence produced from this region. (c) Explain why the mRNA sequence matches one genomic strand exactly, apart from one substitution.
Solution
(a) The gene is on the reverse strand, meaning the reverse strand is the coding (sense) strand — the one whose sequence matches the mRNA. Therefore the forward strand shown here is the template, and reads it 3′→5′.
This is the single most confusing naming convention in molecular biology, so state it carefully: the template strand is the one read by polymerase; the coding strand is the one whose sequence the mRNA reproduces. They are the two strands of the same duplex.
(b) The mRNA is the reverse complement of the shown forward-strand sequence, with U replacing T.
Forward strand (template): 5'-GGCATCGATTACGCAT-3'
Reverse complement: 5'-ATGCGTAATCGATGCC-3'
As RNA: 5'-AUGCGUAAUCGAUGCC-3'
Note it begins with AUG, which is consistent with this being the start of a coding region.
(c) Because the mRNA is synthesised by complementary base pairing against the , and the coding strand is also complementary to the template strand. Two sequences complementary to the same third sequence are identical to each other. The only difference is chemical: RNA uses uracil where DNA uses thymine.
This is why a genome browser can display "the sequence" of a gene without ambiguity once strand is specified — and why forgetting the strand produces a reverse-complemented, biologically meaningless answer.
7. Reading frames
Given the mRNA: 5'-AAAUGCCCGGGUAAUGCUAA-3'
(a) Identify all three reading frames and translate each until a stop . (b) Which frame is most likely the biologically used one, and why? (c) What is the peptide product?
Solution
(a) Write the sequence and split it three ways:
seq: A A A U G C C C G G G U A A U G C U A A
frame 1: AAA UGC CCG GGU AAU GCU AA
Lys Cys Pro Gly Asn Ala ... (no stop reached)
frame 2: AAU GCC CGG GUA AUG CUA A
Asn Ala Arg Val Met Leu ... (no stop reached)
frame 3: AUG CCC GGG UAA UGC UAA
Met Pro Gly STOP (stop at position 10-12)
(b) Frame 3. It is the only frame that begins with an AUG start codon and terminates at an in-frame stop codon — a complete . Frames 1 and 2 simply run off the end of the fragment without either feature.
initiates at an AUG, and in eukaryotes the scans from the to find the first suitable one (Ch 07). The AUG here is at position 3.
(c) Met–Pro–Gly, then termination at UAA.
A tripeptide, which is biologically implausible as a real protein — but the exercise is about frame identification, and the notable feature is the second AUG downstream of the stop. This illustrates why naive ORF-finding is nearly useless in eukaryotic genomic DNA: there are AUGs and stops everywhere by chance, and distinguishing a real coding sequence requires splice signals, codon-usage statistics, and ideally direct transcript evidence (Ch 44).
8. Mutation consequences
The wild-type coding sequence begins: 5'-AUG GCA UUU CGA GGU UCA UAA-3'
For each , give the consequence and its classification.
(a) The 6th base A → G (b) The 7th base U → A (c) The 10th base C → U (d) Deletion of the 8th base (e) ★ The 9th base U → C
Solution
Wild-type, numbered:
pos: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
A U G G C A U U U C G A G G U U C A U A A
Met Ala Phe Arg Gly Ser STOP
(a) Position 6, A→G: 2 GCA → GCG. Both encode Ala. Synonymous (silent) at the protein level. But note the caveat: synonymous is not necessarily consequence-free. It can alter codon usage and rate, mRNA secondary structure, or regulatory elements. Deep-intronic and synonymous variants disrupting splicing are a genuinely under-recognised disease class (Ch 06).
(b) Position 7, U→A: codon 3 UUU → AUU. Phe → Ile. . Both are hydrophobic residues of broadly similar character, so this is a relatively conservative substitution — less likely to be damaging than a change altering charge or size dramatically, though this depends entirely on structural context.
(c) Position 10, C→U: codon 4 CGA → UGA. Arg → STOP. Nonsense. The protein truncates after three residues. In a real multi-exon transcript, a premature stop this far upstream would also trigger , degrading the mRNA rather than producing a truncated protein — so the usual result is via absence of transcript, not a short protein.
(d) Deletion of position 8: everything downstream shifts by one.
new: AUG GCA UUC GAG GUU CAU AA
Met Ala Phe Glu Val His ...
Frameshift. Codon 3 happens to remain Phe (UUU → UUC, both Phe), but every codon after it is different, and the original stop is no longer in frame. Frameshifts are typically severe: they destroy the entire downstream protein sequence and usually introduce a premature stop at some random downstream point.
(e) Position 9, U→C: codon 3 UUU → UUC. Both encode Phe. Synonymous — and specifically a third-position change, which is where the is most degenerate.
This is not a coincidence, and it is the interesting part. The code is arranged so that third-position changes are usually synonymous (), and so that when a substitution does change the amino acid, the replacement tends to be chemically similar. The code has an error-minimising structure — it is far better at buffering mutation and mistranslation than a randomly assigned code would be. That property is almost certainly the product of selection on the code itself.
9. Splicing ★
A gene has three exons. The 1 donor site is mutated from the canonical GU to AU.
(a) What is the immediate molecular consequence? (b) Name three distinct outcomes that could follow. (c) Why might this variant be missed by a standard coding-region analysis?
Solution
(a) The recognises the 5′ splice site largely through base pairing between U1 snRNA and the donor consensus, of which the GU dinucleotide is the most strongly conserved element. Mutating it prevents normal recognition, so intron 1 is not spliced at its correct donor site.
(b) Three genuinely different outcomes:
- Intron retention. The intron is left in the mature mRNA. Introns usually contain in-frame stop , so this typically produces a premature termination codon and triggers .
- Exon skipping. The spliceosome instead joins exon 1's upstream partner to exon 3, removing exon 2 entirely. Whether this is tolerated depends on whether exon 2's length is a multiple of three — if not, the downstream sequence frameshifts.
- Cryptic splice site activation. A nearby sequence that weakly resembles a donor site is used instead, producing a transcript with a slightly shortened or lengthened exon, again with frame consequences depending on the shift.
Which occurs is often unpredictable from sequence alone, and more than one can occur in the same tissue, producing a mixture of transcripts.
(c) Because the variant is not in the coding sequence. It sits at the exon–intron boundary, in what a naive coding-region treats as intronic and therefore filters out. Analyses that consider only and nonsense changes within exons will discard it.
This generalises: splice-disrupting variants — including deep-intronic ones creating new cryptic sites, and synonymous exonic ones destroying — are a substantial and historically under-ascertained cause of Mendelian disease. It is one of the main reasons leaves cases undiagnosed that genome sequencing plus RNA analysis can solve (Ch 54).
10. Packing arithmetic
The human genome is about 6.2 × 10⁹ bp. B-DNA rises 0.34 nm per base pair. A typical nucleus is about 6 μm across.
(a) What is the total contour length of DNA in one diploid cell? (b) What linear compaction factor is required to fit it into the nucleus? (c) A wraps 147 bp in about 1.7 turns, occupying roughly 11 nm of linear space where the naked DNA would span 50 nm. What compaction does the nucleosome alone achieve, and what does that tell you?
Solution
(a) 6.2 × 10⁹ bp × 0.34 nm/bp = 2.108 × 10⁹ nm = ~2.1 metres
Two metres of DNA in every nucleated cell in your body.
(b) Comparing end-to-end length with nuclear diameter:
2.1 × 10⁹ nm ÷ 6 × 10³ nm = ~350,000-fold linear compaction
But this comparison is the wrong one, and recognising why is the real content of the problem. DNA is not a rigid rod that must be shortened; it is a flexible polymer whose unconstrained size is set by random-coil statistics, not contour length. The genuine problem is not fitting 2 m into 6 μm — it is imposing an ordered, untangled, locally accessible conformation on a polymer that would otherwise adopt a knotted random coil.
(c) Nucleosome compaction = 50 nm ÷ 11 nm ≈ 4.5-fold (commonly quoted as ~6-7 fold depending on how linker DNA is counted).
What that tells you: the nucleosome achieves only a tiny fraction of the total compaction needed. Even the 30 nm fibre — whose existence in vivo is not established (Ch 03) — reaches only ~40-fold cumulative. The remaining several orders of magnitude come from , compartmentalisation and large-scale folding (Ch 50), not from a tidy hierarchy of coils within coils.
The textbook ladder of 10 nm → 30 nm → 300 nm loops → is a diagram, not a finding.
Where you went wrong
Track your errors by type — the patterns repeat far more than the content does.
| Error pattern | What to re-read |
|---|---|
| Lost track of 5′/3′ or strand | Ch 02, and problem 1(d) |
| Confused template with coding strand | Ch 05, problem 6 |
| Treated synonymous as "no effect" | Ch 07, problem 8(a) |
| Forgot frameshift destroys everything downstream | Problem 8(d) |
| Assumed one mechanism gives the final fidelity | Ch 04, problem 5 |
| Compared contour length to nuclear diameter without questioning it | Problem 10(b) |