Problem set 06 — Gene regulation
Covers Ch 21–25.
Attempt before revealing. problems are fully determined by an algorithm, so reading the worked answer teaches you nothing. Run it yourself, on paper, every time.
Notation. I is lacI (), P the lac , O the operator, Z lacZ (β-galactosidase), Y lacY (permease). P⁻ breaks the lac promoter only — lacI has its own promoter. A slash separates (left) from F′ (right). ★ marks the two hardest.
1. The lac truth table
(a) Fill in the four rows: lactose present/absent × glucose present/absent, giving operator state, CRP state, and output. (b) Name the actual molecular signal carrying each input. Neither is the sugar in the column heading. (c) Write the Boolean expression, then test it. Using the occupancy model of Ch 21 §4 — rate ∝ f_free × A with A = 0.03 + 0.97·p_C, f_free = 0.001 / 0.15 / 0.95 and p_C = 0.05 (glucose present) or 0.90 (absent) — compute the four rates. Is this an AND gate?
Solution
(a)
| Lactose | Glucose | Operator | CRP–cAMP | lacZYA |
|---|---|---|---|---|
| − | + | occupied by LacI | absent | off |
| − | − | occupied by LacI | present | off |
| + | + | mostly occupied (inducer exclusion) | absent | barely on |
| + | − | free | present | ON |
(b) Trap, and the commonest error on this .
- The lactose input is carried by , an isomer made as a side reaction by β-galactosidase — the enzyme this operon encodes. Lactose itself binds nothing regulatory.
- The glucose input is carried by the act of transporting glucose. Import leaves EIIAGlc unphosphorylated, which (i) fails to activate adenylate cyclase, so cAMP falls and CRP loses its ligand, and (ii) binds LacY directly, shutting the permease (inducer exclusion).
So the "lactose" row really means "allolactose present", which requires pre-existing permease and pre-existing β-galactosidase. That is why repression must be leaky: a perfect makes the operon permanently uninducible.
(c) expression = allolactose AND NOT glucose-transport. Compute A:
- glucose present: A = 0.03 + 0.97 × 0.05 = 0.0785
- glucose absent: A = 0.03 + 0.97 × 0.90 = 0.903
| Lactose | Glucose | f_free | A | rate = f_free × A | normalised |
|---|---|---|---|---|---|
| − | + | 0.001 | 0.0785 | 7.85 × 10⁻⁵ | 0.000092 |
| − | − | 0.001 | 0.903 | 9.03 × 10⁻⁴ | 0.00105 |
| + | + | 0.15 | 0.0785 | 1.178 × 10⁻² | 0.0137 |
| + | − | 0.95 | 0.903 | 8.579 × 10⁻¹ | 1.000 |
Dynamic range = 8.579 × 10⁻¹ / 7.85 × 10⁻⁵ = 10,930-fold. Switch-like, certainly.
But it is not an AND gate: the three "off" states are not equal. The loudest exceeds the quietest by 1.178 × 10⁻² / 7.85 × 10⁻⁵ = 150-fold — exactly the f_free ratio 0.15 / 0.001, since both states are glucose-present and A cancels. A digital AND has one 0; this has three, spanning two orders of magnitude. What makes it look digital is positive feedback, which gives bistability — so the population mean reports a value no single cell holds (Ch 48).
2. lacI⁻ — a broken trans factor
Predict β-galactosidase, with and without inducer (IPTG).
| # | − IPTG | + IPTG | |
|---|---|---|---|
| 1 | I⁺ O⁺ Z⁺ | ||
| 2 | I⁻ O⁺ Z⁺ | ||
| 3 | I⁻ O⁺ Z⁺ / F′ I⁺ O⁺ Z⁻ | ||
| 4 | I⁺ O⁺ Z⁻ / F′ I⁻ O⁺ Z⁺ |
(a) Fill in the table. (b) Rows 3 and 4 put the same on opposite molecules. Why does that not matter? (c) LacI is present at ~10 tetramers per cell, and each lac copy carries three operators (O1, O2, O3), so a has two copies. Is one functional lacI enough? Show the arithmetic.
Solution
(a)
| # | Genotype | − IPTG | + IPTG | |
|---|---|---|---|---|
| 1 | I⁺ O⁺ Z⁺ | − | + | inducible (wild type) |
| 2 | I⁻ O⁺ Z⁺ | + | + | constitutive |
| 3 | I⁻ O⁺ Z⁺ / F′ I⁺ O⁺ Z⁻ | − | + | inducible — rescued |
| 4 | I⁺ O⁺ Z⁻ / F′ I⁻ O⁺ Z⁺ | − | + | inducible — rescued |
Row 3 by the algorithm: pool the trans products — the F′ I⁺ puts functional LacI in the cell and it diffuses to both operators. O⁺ repressed, Z⁺ the only enzyme source; F′ O⁺ repressed, Z⁻ silent anyway. Net: inducible.
(b) Because LacI is a protein, and a protein leaves the DNA that encoded it. Which molecule carried the gene is invisible to the assay. That is what trans-acting means, and why lacI⁻ is : one good copy anywhere supplies the missing global.
Trap. "Recessive" here is not inferred from a cross — no , no gametes, no 3:1. The merodiploid test asks exactly one question: does a second copy fix it? Yes → the lesion is in a diffusible product. No → the lesion is in the DNA itself.
(c) Sites to cover: 3 operators × 2 copies = 6. Capacity: a LacI tetramer is a dimer of dimers, so each grips two operators (that bridging is what loops the DNA):
10 tetramers × 2 sites = 20 site-equivalents against 6 sites — a threefold excess.
One functional lacI is sufficient, which is what the experiment shows. Note how tight it is: repression with all three operators is ~1,300-fold, so O1 is unoccupied a fraction 1/1,300 = 7.7 × 10⁻⁴ of the time — the same order of magnitude as the round f_free = 0.001 problem 1 uses. (Those occupancies are their own model, not this measurement: 0.95 / 0.001 is 950-fold repression, not 1,300. Same ballpark, different provenance.)
3. lacIˢ — a trans factor that poisons
lacIˢ binds the operator normally but cannot be released by inducer.
| # | − IPTG | + IPTG | |
|---|---|---|---|
| 1 | Iˢ O⁺ Z⁺ | ||
| 2 | Iˢ O⁺ Z⁺ / F′ I⁺ O⁺ Z⁺ | ||
| 3 | I⁺ O⁺ Z⁺ / F′ Iˢ O⁺ Z⁻ | ||
| 4 | Iˢ O⁺ Z⁻ / F′ I⁻ O⁺ Z⁺ |
(a) Fill in the table. (b) Row 2 is . Give the mechanism, and explain why dominance here is not evidence that the mutant protein is more active. (c) If Iˢ and I⁺ subunits are made in equal amounts and tetramers assemble at random, what fraction is pure wild type? With ~10 tetramers per cell, how many does a cell hold? (d) Repeat (c) with mutant subunits at 10% of the pool. What does the comparison say about how absolute dominance is?
Solution
(a) All four are uninducible — no β-galactosidase under any condition.
| # | Genotype | − IPTG | + IPTG |
|---|---|---|---|
| 1 | Iˢ O⁺ Z⁺ | − | − |
| 2 | Iˢ O⁺ Z⁺ / F′ I⁺ O⁺ Z⁺ | − | − |
| 3 | I⁺ O⁺ Z⁺ / F′ Iˢ O⁺ Z⁻ | − | − |
| 4 | Iˢ O⁺ Z⁻ / F′ I⁻ O⁺ Z⁺ | − | − |
Row 3 is the one to check. Iˢ sits on the F′, whose Z⁻ makes no enzyme — yet the strain is uninducible, because the superrepressor diffuses to the chromosomal operator and locks the intact Z⁺ copy off. Which molecule carries a trans factor is irrelevant. Hold that until problem 4, where the opposite holds. Row 4: the pool is pure Iˢ, locking the F′'s O⁺.
(b) LacI is a homotetramer, so wild-type and mutant subunits co-assemble into mixed tetramers. A mixed tetramer still binds DNA but cannot be released, because release requires the whole particle to make the allosteric shift. A minority of mutant subunits disables the majority — negative complementation: dominance from multimerisation, not hyperactivity.
Trap. Dominance describes how a lesion behaves in the presence of a good copy, not how strong the mutant product is. Iˢ has lost a function yet is dominant, while I⁻, which has lost more function, is . Loss-of-function does not imply recessive.
(c) Four subunits drawn independently from a 50:50 pool:
P(all four wild type) = (½)⁴ = 1/16 = 6.25%; P(≥ 1 mutant) = 93.75% Expected pure-wild-type tetramers = 10 × 0.0625 = 0.625 per cell
A typical cell holds zero functional particles, so at this copy number the dominance is close to absolute.
(d) Mutant subunits at 10%:
P(all four wild type) = 0.9⁴ = 65.6%; P(poisoned) = 34.4% Expected pure-wild-type tetramers = 10 × 0.6561 = 6.6 per cell — substantially inducible.
Dominance is quantitative and dose-dependent, not a binary label, and the exponent 4 amplifies it: 10% → 50% mutant subunits takes the poisoned fraction from 34% to 94%.
4. lacOᶜ — a cis element, and why "dominant" is the wrong word
lacOᶜ is an operator sequence no can bind.
| # | − IPTG | + IPTG | |
|---|---|---|---|
| 1 | I⁺ Oᶜ Z⁺ | ||
| 2 | I⁺ Oᶜ Z⁺ / F′ I⁺ O⁺ Z⁺ | ||
| 3 | I⁺ Oᶜ Z⁻ / F′ I⁺ O⁺ Z⁺ | ||
| 4 | I⁻ Oᶜ Z⁻ / F′ I⁺ O⁺ Z⁺ |
(a) Fill in the table. (b) Rows 2 and 3 differ by one — Z⁺ versus Z⁻ on the Oᶜ copy. Why does that change the completely? (c) For row 2, predict the quantitative ratio of uninduced to induced β-galactosidase, taking full repression as 1,300-fold. Why is the ratio diagnostic?
Solution
(a)
| # | Genotype | − IPTG | + IPTG | Phenotype |
|---|---|---|---|---|
| 1 | I⁺ Oᶜ Z⁺ | + | + | constitutive |
| 2 | I⁺ Oᶜ Z⁺ / F′ I⁺ O⁺ Z⁺ | + | ++ | constitutive |
| 3 | I⁺ Oᶜ Z⁻ / F′ I⁺ O⁺ Z⁺ | − | + | inducible |
| 4 | I⁻ Oᶜ Z⁻ / F′ I⁺ O⁺ Z⁺ | − | + | inducible |
Row 3: the trans pool is healthy. The chromosomal copy is Oᶜ, so it ignores the repressor and is transcribed constitutively — but its Z⁻ makes that constitutivity invisible to the assay, and the F′ O⁺ Z⁺ supplies all the enzyme, inducibly. Row 4 repeats the logic with two decoys: I⁻ rescued in trans, Oᶜ stranded on a Z⁻ copy. Three , none detectable.
(b) Because an operator is a sequence, not a product: it can only affect of genes on its own DNA molecule, so it is scored on that molecule's structural genes and nowhere else.
Trap, and the point of this problem set. Oᶜ is called "cis-", and dominant misleads. It is not out-competing O⁺; it is not interacting with it at all, and has literally zero effect on the other copy. Whether you can detect it depends on which structural genes sit downstream of it on its own molecule — a fact about your reporter, not about dominance. Contrast problem 3 row 3, where a trans factor on a Z⁻ molecule was fully detectable.
(c) Take one fully derepressed copy as 1 unit; a repressed O⁺ copy runs at 1/1,300 = 7.69 × 10⁻⁴.
Uninduced = 1 (the Oᶜ copy) + 7.69 × 10⁻⁴ = 1.00077 Induced = 1 + 1 = 2; ratio = 1.00077 / 2 = 0.500
The strain expresses ~50% of its induced level with no inducer; the haploid of row 1 gives 1/1 = 100%. That is diagnostic — a fully constitutive strain has no repressible copy, a half constitutive strain has exactly one. Measuring the ratio rather than scoring ± tells you the copy number and lets you infer the genotype. This is "partial constitutivity": a quantitative signature, not a hedge.
5. ★ Separating cis from trans when both are broken
Both reporters assayed now. Predict β-galactosidase and permease, ± IPTG.
| # | |
|---|---|
| a | Iˢ P⁺ O⁺ Z⁺ Y⁻ / F′ I⁺ P⁺ Oᶜ Z⁻ Y⁺ |
| b | I⁻ P⁻ Oᶜ Z⁺ Y⁺ / F′ I⁺ P⁺ O⁺ Z⁻ Y⁺ |
| c | I⁺ P⁺ O⁺ Z⁻ Y⁺ / F′ I⁻ P⁺ Oᶜ Z⁺ Y⁻ |
| d | Iˢ P⁻ O⁺ Z⁺ Y⁺ / F′ I⁺ P⁺ Oᶜ Z⁺ Y⁻ |
(a)–(d) Work each one. (e) You isolate a new constitutive mutant, genotype ? Z⁺. Design the single that distinguishes I⁻ from Oᶜ. State the exact F′ and why.
Solution
The algorithm: (1) pool the trans products; (2) take each copy separately, first, then operator; (3) check that copy's structural genes; (4) sum.
(a) Pool: Iˢ → mixed tetramers → locked onto DNA, inducer irrelevant. P⁺ O⁺ → locked off, so Z⁺ is never expressed and Y⁻ is broken anyway. F′ P⁺ Oᶜ → immune → constitutive; Z⁻ no enzyme, Y⁺ yes.
(b) Pool: F′ I⁺ → normal inducible repressor. Chromosome P⁻ → no of anything, ever; the Oᶜ sits downstream of a dead promoter and is meaningless, Z⁺ Y⁺ silent. F′ P⁺ O⁺ → inducible; Z⁻ none, Y⁺ inducible.
Trap. Oᶜ here is not merely invisible to one reporter — it is mechanistically inert. You cannot regulate a transcript that is never made. Check the promoter before the operator.
(c) Pool: chromosomal I⁺ → normal repressor. Chromosome P⁺ O⁺ → inducible; Z⁻ none, Y⁺ inducible permease. F′ Oᶜ → constitutive; Z⁺ constitutive β-gal, Y⁻ none.
(d) Pool: Iˢ → locked. Chromosome P⁻ → nothing, so O⁺ Z⁺ Y⁺ all silent. F′ Oᶜ → immune → constitutive; Z⁺ yes, Y⁻ none.
| β-gal −IPTG | β-gal +IPTG | Permease −IPTG | Permease +IPTG | |
|---|---|---|---|---|
| a | none | none | + | + |
| b | none | none | − | + |
| c | + | + | − | + |
| d | + | + | none | none |
Row c is the one to memorise: one cell, two genes of the same , one constitutive and one inducible. There is no such thing as a "constitutive cell" — constitutivity is a property of a copy, and the assay reports whichever copy carries the reporter.
Row d shows that a trans- lesion and a cis-dominant lesion in one cell do not fight. Iˢ wins wherever repressor can bind; Oᶜ is a molecule where it cannot.
(e) Introduce F′ I⁺ O⁺ Z⁻.
- Lesion is
I⁻: the F′ supplies repressor in trans, it binds the chromosomalO⁺, and the strain becomes inducible. - Lesion is
Oᶜ: the chromosomal operator binds no repressor however much arrives, so the strain stays constitutive.
The Z⁻ is the load-bearing design choice and the part people omit: it guarantees all measured β-galactosidase comes from the chromosome, making the readout binary. With F′ I⁺ O⁺ Z⁺ you would still get an answer — I⁻ → inducible, Oᶜ → ~50% constitutive by problem 4(c) — but you would be discriminating two levels of a continuous quantity instead of presence from absence.
6. trp attenuation
The trp leader encodes a 14- peptide with Trp codons at positions 10 and 11. Four RNA segments: 3:4 is an intrinsic terminator, 2:3 an antiterminator that consumes segment 3.
Model TrpR as 70-fold control of initiation and as 9-fold control of readthrough (readthrough 0.90 when Trp is scarce, 0.10 when plentiful).
(a) For high and for low tryptophan, state where the sits, which hairpin forms, and whether polymerase reaches trpE. (b) Compute expression in both conditions and the combined fold-control. (c) A deletes the leader start codon, so no ribosome ever loads. Predict the outcome under tryptophan starvation. (d) A mutation changes both leader Trp codons to Arg codons. Predict expression under starvation, and how much regulation survives. (e) Why can no eukaryote use this mechanism? Be precise about what is forbidden.
Solution
(a) The measurement is the ribosome's speed; its position is the readout.
Trp plentiful — charged tRNA-Trp abundant, the ribosome sails through codons 10–11, reaches the leader stop codon and rests covering segment 2. With 2 sequestered, 2:3 cannot form; 3:4 pairs into the terminator and polymerase releases at the U-tract. trpEDCBA not transcribed.
Trp scarce — uncharged tRNA-Trp, the ribosome stalls at codons 10–11 on segment 1, leaving 2 free. 2:3 forms, consuming segment 3 so 3:4 cannot. Polymerase reads through.
Two consecutive Trp codons in a 14-codon peptide is a statistical scream — tryptophan is the rarest amino acid, ~1% of residues. The peptide is an instrument, not a product.
(b) Take fully derepressed initiation as 1.
Low Trp: 1 (TrpR is apo and cannot bind DNA) × 0.90 = 0.90 High Trp: 1/70 = 0.01429 (Trp is a corepressor — it puts TrpR on the DNA) × 0.10 = 1.429 × 10⁻³
Combined = 0.90 / 1.429 × 10⁻³ = 630-fold = 70 × 9. The mechanisms act at different control points (initiation, elongation), so they multiply.
(c) With no ribosome at all, 1:2 and 3:4 both pair: the terminator forms and stops, even though tryptophan is absent and the enzymes are needed. This reveals the design — the default is terminate, and it takes an actively stalled ribosome to switch readthrough on. The system does not detect missing tryptophan; it detects a ribosome in trouble, so anything that stops leader reads as "plenty of tryptophan".
(d) With Arg codons at 10–11 the ribosome no longer stalls when Trp is scarce; Arg is abundant, so it parks on segment 2 exactly as in the plentiful case.
Under starvation: 1 × 0.10 = 0.10, i.e. 0.90/0.10 = 9-fold below wild type, precisely when the cell needs the enzymes most.
Surviving range: high Trp gives (1/70) × 0.10 = 1.429 × 10⁻³, so 0.10 / 1.429 × 10⁻³ = 70-fold — TrpR alone. Attenuation is not broken; it is decoupled from the signal. It still fires, always the same way. A sensor stuck at one reading is worse than no sensor, because it still costs.
(e) It requires a ribosome translating a transcript while polymerase is still making it, so ribosome position determines which hairpin forms ahead of the polymerase. Bacteria have no nucleus; in a eukaryote the transcript is capped, spliced and exported before any ribosome touches it.
Be precise about what is forbidden. Attenuation in the general sense — a leader RNA deciding whether an already-initiated transcript continues — thrives in eukaryotes: -proximal Pol II termination is genome-wide, and eukaryotic TPP make that decision by RNA folding. Only the ribosome-as-sensor version is ruled out. The nuclear envelope does not add a processing step; it deletes a class of mechanism.
7. Three ways to break a eukaryotic locus
ENH-L is a limb-bud-specific ; geneA has separate enhancers active in gut and brain; geneB, in the neighbouring domain, is dosage-sensitive and normally brain-only.
CTCF-B1 ▶ ◀ CTCF-B2
├───────────────── TAD A (850 kb) ──────────────────┤ ├── TAD B ──┤
[ENH-L]────────── 350 kb ──────────[P_geneA]─[geneA] [P_geneB]─[geneB]
Predict each consequence, naming which genes and which tissues:
(a) A 600 bp deletion removing ENH-L. (b) A point destroying the in P_geneA. (c) A 3 kb deletion removing CTCF-B2. (d) Which would detect? Which are cis-acting? (e) ENCODE lists 2,348,854 candidate cis-regulatory elements against 19,442 protein-coding genes in a 3.1 × 10⁹ bp genome. Compute elements per gene, mean spacing, and how many candidates sit inside A.
Solution
(a) Enhancer deletion — tissue-specific loss, or nothing. geneA loses limb-bud expression; gut and brain are untouched, driven by different enhancers acting on the same . geneB unaffected. This is the classic non-coding disease mechanism, and why a limb malformation can be caused by a variant 350 kb from any coding base.
Trap. The honest prediction has a second branch: frequently, deleting an enhancer does nothing measurable, because shadow/redundant enhancers drive the same pattern. Redundancy is the norm in development, so "no " is weak evidence that an element is non-functional.
(b) Promoter mutation — all tissues, one gene, one . The core promoter is the common output point every enhancer must act through, so geneA is degraded in limb, gut and brain alike; geneB unaffected. That is the sharpest contrast with (a): enhancer lesions are tissue-restricted, promoter lesions are not. Mechanistically, enhancers set burst frequency and core promoter elements set burst size, so this reduces how much RNA each firing event makes rather than how often the gene fires. (Only ~10% of human promoters carry a TATA box, so this is not a generic lesion.)
(c) Boundary deletion — a gain of expression, in the wrong gene. geneA keeps its own enhancers and is roughly fine. The consequence is that ENH-L, no longer confined, can contact P_geneB and drive geneB ectopically in the limb bud — enhancer hijacking, as at the EPHA4 boundary in limb malformation and at boundaries removed by in cancer.
Trap. The instinct is , because deletions usually cause loss. Here a deletion causes a gain, of the wrong gene, in the wrong tissue, and the affected gene is not the one nearest the deletion. Deleting a is not the same as deleting a function.
(d) Exome sequencing detects none of them — 350 kb upstream, a core promoter, an insulator. All three are cis-acting, which is the payoff of problems 2–5: each affects only the it sits on, so in a heterozygote each produces allele-specific expression. That is the test run in a human, with the homologous chromosome playing the F′ (Ch 52).
(e) Elements per protein-coding gene: 2,348,854 / 19,442 = 120.8
Mean spacing: 3.1 × 10⁹ / 2,348,854 = 1,320 bp — a candidate element every ~1.3 kb
Candidates inside TAD A: 850,000 / 1,320 = 644
A non-coding variant in this TAD sits among ~644 candidates, of which perhaps one is causal — a prior of 1/644 ≈ 0.16% before any other evidence, which is why nearest-gene assignment fails so often. And 2.3 million counts candidates: an upper bound on real regulation, a lower bound on the interpretation problem.
8. ★ Feed-forward loops: what each one computes
X is an input, Y an intermediate TF, Z the output.
- Coherent type-1 (C1-FFL), AND logic: X → Y; X and Y → Z.
- Incoherent type-1 (I1-FFL): X → Z; X → Y ⊣ Z.
In both, dY/dt = β − αY while X is on with β = 10 units/min and α = 0.1 min⁻¹, and dY/dt = −αY when X is off; Y starts at 0. For the C1-FFL, Z fires iff X is present and Y ≥ 50. For the I1-FFL, take Z as fast relative to Y, so Z(t) = 100 × 30/(30 + Y(t)) while X is on.
(a) Write Y(t); compute the steady state and the C1-FFL's ON-delay. (b) Give a 4-minute pulse of X. Does Z fire? Show the number. (c) Give a 30-minute sustained X, removed at t = 30. When does Z switch on and off? Explain the asymmetry. (d) For the I1-FFL compute Z at t = 0, 5, 10, 20 and at steady state. Give the peak-to-steady ratio and the time at which Z falls to half its peak. (e) Name the computation each performs, and say which you would wire to an expensive irreversible response driven by a noisy input. (f) Rebuild the C1-FFL with OR logic. Where does the delay move? Compute it.
Solution
(a) Y(t) = (β/α)(1 − e^(−αt)) = 100(1 − e^(−0.1t)), steady state Y* = 10/0.1 = 100. Note what sets the timescale: α alone. Raising β raises the final level but does not make it arrive sooner.
ON-delay — Y reaching 50, which is half of Y*:
100(1 − e^(−0.1t)) = 50 → e^(−0.1t) = 0.5 → t = ln 2 / 0.1 = 6.93 min
The threshold sits at half the steady state, so the delay is exactly one protein half-life.
(b) Y(4) = 100(1 − e^(−0.4)) = 100(1 − 0.67032) = 32.97 units
32.97 < 50, so Z never fires, and the AND fails outright the moment X leaves. Minimum pulse that fires Z: 6.93 min.
(c) Z switches on at t = 6.93 min and stays on while X persists; Y(30) = 100(1 − e⁻³) = 100(1 − 0.049787) = 95.02 units, output window 30 − 6.93 = 23.07 min. Z switches off at t = 30, instantly — the AND needs X, and X is gone; Y is still at 95 and irrelevant.
The asymmetry is the circuit. Delay on ON, none on OFF: a debounce. Transients shorter than 6.93 min are discarded, sustained signals get through, withdrawal is obeyed at once.
(d) Z(t) = 3000/(30 + Y(t)):
| t (min) | Y(t) | 30 + Y | Z(t) |
|---|---|---|---|
| 0 | 0 | 30 | 100.0 |
| 5 | 39.35 | 69.35 | 43.3 |
| 10 | 63.21 | 93.21 | 32.2 |
| 20 | 86.47 | 116.47 | 25.8 |
| ∞ | 100 | 130 | 23.1 |
(Working: e^(−0.5) = 0.60653, e^(−1) = 0.36788, e^(−2) = 0.13534; steady state 3000/130 = 23.08.)
Peak-to-steady = 100 / 23.08 = 4.33-fold
Half-peak (Z = 50): 3000/(30 + Y) = 50 → Y = 30 → 100(1 − e^(−0.1t)) = 30 → e^(−0.1t) = 0.7 → t = ln(1/0.7)/0.1 = 3.57 min
Z peaks immediately and is half-gone within 3.6 minutes while X is still fully on. A pulse, generated from a step.
Trap. Nothing about the input is transient — X is on throughout. The transience is manufactured by the circuit, so observing a pulse of expression does not license concluding that the signal was a pulse.
(e) The C1-FFL with AND is a persistence detector — "has X been on long enough to mean something?" The I1-FFL is a pulse generator, plus an accelerated rise and, tuned appropriately, fold-change detection: it responds to the ratio by which X changed rather than its absolute level, answering "did something just change?" Many animal I1-FFLs use a microRNA as the repressing arm (Ch 24).
For an expensive irreversible response to a noisy input: the coherent AND FFL. Committing to an irreversible fate on a 4-minute fluctuation is the exact failure the delay prevents, and a pulse generator is wrong twice over — it fires on the transient, then shuts off regardless.
(f) OR logic moves the delay to the OFF edge. X alone satisfies the OR, so there is no ON-delay; Z fires at t = 0. When X is withdrawn at t = 30, Y = 95.02 and sustains Z alone until it decays below 50:
95.02 · e^(−0.1Δt) = 50 → e^(−0.1Δt) = 50/95.02 = 0.52620 → −0.1Δt = ln(0.52620) = −0.64207 → Δt = 6.42 min
Z persists 6.42 min after X disappears — a persistence detector on the loss of signal, the right design when interruption is expensive and restarting is slow.
Same three nodes, same edges, same signs; changing only the logic at the Z inverts which edge is filtered. The wiring diagram does not determine the computation — the wiring diagram plus the input function does, which is why motif-counting tells you where to look, not what you will find.
Where you went wrong
| Error pattern | What to re-read |
|---|---|
| Said lactose is the inducer, or that glucose binds something | Problem 1(b) |
| Treated lac as a genuine digital AND gate | Problem 1(c) — off-states differ 150-fold |
| Asked which molecule a trans factor sits on | Problem 3(a) row 3 — it never matters |
| Assumed loss-of-function implies | Problem 3(b) — dominance by multimerisation |
Scored Oᶜ without checking which genes are downstream of it | Problems 4(b) and 5(c) |
| Read "cis-" as Mendelian dominance | Problem 4(b) — Oᶜ never touches the other copy |
| Checked the operator before the | Problem 5(b) — P⁻ makes everything downstream moot |
| Called a whole cell "constitutive" | Problem 5(c) — one , one gene each way |
| Thought senses tryptophan | Problem 6(c) — no reads as "plenty" |
| Predicted from a boundary deletion | Problem 7(c) — it causes ectopic gain |
| Assigned a non-coding variant to the nearest gene | Problem 7(e) — ~644 candidates per |
| Inferred a transient signal from a transient response | Problem 8(d) — the circuit makes the pulse |
| Read a motif's wiring and stopped there | Problem 8(f) — AND and OR give opposite filters |