What Zifo usually asks
Pieced together from candidate reports on Glassdoor and GeeksforGeeks (2016–2026). These are candidate reports, not official, and the details vary by campus and year.
Aptitude
"Simple" quant + logical, with negative marks. Topics reported: ratios, ages, speed–distance, probability, Venn diagrams, syllogisms, some English and statistics. One past report: 40 Qs in 45 min, so expect about 50 s per question.
Science
MCQs from the topic you chose in advance. Stoichiometry and drug development are both mentioned. Expect mostly direct concepts and quick numericals, not deep theory.
Essay
Candidates report the essay is based on a picture shown during the test. You describe it, interpret it, and give your view. Prepare for that format, not just topic essays.
Gamified
Reported: puzzles over multiple levels, multitasking, coded diagrams, and moving a ball into a hole (a tilt or route puzzle). Levels get harder as you go.
After the OA, candidates report a personal/HR interview: self-intro, resume and projects, logical puzzles you solve out loud, and "Why Zifo?" See the
Interview section.
Aptitude cheat sheet
Know these formulas cold so each question takes one step. Then practise with the 30-question quiz.
Negative marking: when to guess
Say a correct answer gives +1, a wrong one −p, and a skip 0. A guess pays off on average only if your chance of being right is greater than
p / (1+p).
- −¼ penalty, 4 options: a blind guess has EV = 0, so don't. Eliminate 1 option (33% chance) and EV turns positive, so guess.
- −⅓ penalty: you need more than 25%. Eliminate 1 option of 4 first.
- −1 penalty: you need more than 50%. Answer only if you're down to 2 options and lean clearly one way.
Check the penalty on the instruction screen and apply the matching rule.
Three-pass time strategy (35 min)
- Pass 1 (~20 min): answer everything you can solve in under 40 s with confidence. Mark anything longer for review and move on.
- Pass 2 (~12 min): go back to the marked questions you know how to solve.
- Pass 3 (last 3 min): make guesses only where the rule above says so, then stop. Never blind-guess to "fill" answers.
- If you're still stuck at 60 s, skip. One hard question isn't worth three easy ones.
- Read the last line of the question first, so you know exactly what's asked (ratio? %? which unit?).
- Use the options: plug them back in (age, work and number questions) and estimate to eliminate.
Quant formulas
| Topic | Must-know |
|---|
| Percentages | Successive changes a% and b%: net = a+b+ab/100. If A is r% more than B, then B is r/(100+r)×100% less than A. Fractions: 1/8 = 12.5%, 1/6 = 16.67%, 1/7 ≈ 14.28%, 1/9 ≈ 11.11%, 1/11 ≈ 9.09%. |
| Profit & loss | Profit% = P/CP×100. SP = CP(1+P%/100). Marked price → SP after discount d%: MP(1−d/100). Successive discounts work like successive % changes. False weight gain% = error/(true−error)×100. |
| Interest | SI = PRT/100. CI: A = P(1+R/100)^T. CI−SI over 2 yrs = P(R/100)². A sum doubles under SI in 100/R years. |
| Ratio & ages | Split a total in ratio a:b → a/(a+b)×Total. For ages, write "x years ago / hence" equations with a single variable. The age difference never changes. |
| Averages | Sum = avg × n. Removed value = old sum − new sum. If a newcomer changes the avg by d: newcomer = old avg + (n+1)·d. |
| Mixtures | Alligation: (dearer−mean) : (mean−cheaper) = cheaper qty : dearer qty. Repeated replacement: left = x(1−y/x)^n. |
| Speed–distance | km/h × 5/18 = m/s. Avg speed for equal distances = 2xy/(x+y). Relative speed: add if opposite directions, subtract if same. Train passing a pole: its own length. Passing a platform: train + platform. Boats: down = b+s, up = b−s, so s = (down−up)/2. |
| Time & work | Together = ab/(a+b) days. Use LCM as the total work. Pipes: filling pipes are +, leaking pipes are −. |
| Number system | Divisibility: 3/9 (digit sum), 4 (last 2 digits), 8 (last 3), 11 (alternating sum). HCF×LCM = a×b (for two numbers). Unit digit cycles: 2,3,7,8 cycle every 4; 4,9 every 2; 0,1,5,6 never change. |
| P & C | nPr = n!/(n−r)!, nCr = n!/r!(n−r)!. Circular arrangements: (n−1)!. Repeated letters: n!/(p!q!). "At least one" = total − none. |
| Probability | P = favourable/total. P(not A) = 1−P(A). Independent events: multiply. P(A∪B) = P(A)+P(B)−P(A∩B). 2 dice: 36 outcomes (sum 7 → 6 ways, sum 8 → 5). Deck: 52 cards, 4 suits, 12 face cards, 4 aces. |
| Clocks / Calendar | Angle = |30H − 5.5M|. Hands overlap every 65 5/11 min. Odd days: normal yr 1, leap yr 2. Day after N days: shift by N mod 7. |
| Basic stats | Mean, median (middle value of sorted data), mode. Range. Variance = mean of squared deviations. Adding a constant changes the mean, not the SD. Multiplying by k scales both. |
Logical reasoning
Syllogisms: draw Venn circles. "All A are B" + "Some B are C" → no definite conclusion about A and C. If two conclusions are complementary ("Some A are C" / "No A is C"), the answer is "either". Only accept a conclusion if it holds in every diagram. "All A are B" also implies "Some B are A".
Venn counting: n(A∪B) = A + B − A∩B. Neither = Total − n(A∪B). For 3 sets, add back the triple overlap. Fill the diagram from the centre outward.
Blood relations: draw a family tree (+ male, − female, = married). "Only son of my grandfather" = father.
Directions: sketch N up. Right turn from N → E. Use Pythagoras for straight-line distance. Shadows: morning sun in the E, so the shadow falls W.
Coding–decoding: check letter shifts (+1, +2), reversals, and positions (A=1…Z=26; reverse value = 27−n). Remember EJOTY = 5, 10, 15, 20, 25.
Series: try differences, then second differences, then ×/÷, squares/cubes (n²+n, n³), primes, or two interleaved series.
Seating/arrangement: draw the slots first, place the fixed clues, then branch on "either/or" clues. Circle facing centre: left = clockwise.
Data sufficiency: test statement 1 alone, then 2 alone, then both together. You don't need the actual value, only whether one can be found.
Verbal (if it appears)
- Error spotting: subject–verb agreement ("Each of the boys is"; "Neither…nor" agrees with the nearer subject), tense consistency, articles, prepositions (superior to, prefer X to Y, comprises with no "of"), parallelism.
- Reading comprehension: read the questions first. Look for the answer in the passage, not in your own opinion. Extreme options ("always", "never") are usually wrong.
- Para-jumbles: find the opening sentence (no pronouns or connectors) and lock in pairs like "This…" following its referent.
- Vocabulary: meticulous (careful), ephemeral (short-lived), pragmatic (practical), ambiguous (unclear), candid (frank), diligent (hard-working), obsolete (outdated), lucid (clear), mitigate (reduce), verbose (wordy).
Science section: all 6 topics
You only need one topic. Pick it, master its notes, and do its quiz until you score 90%+.
Which to pick (10 min, negative marking):
- Stoichiometry: best default. About 8 formulas, exact answers, little to memorise.
- Chromatography: a good second choice. A handful of formulas (k, N, Rs) plus detector facts.
- Statistical Inference: choose it if you're comfortable with stats or probability.
- Drug Development / Clinical Trials: choose these if you have a pharma or biotech background. They're mostly terms and phase facts.
- Drug Discovery: choose it if you know pharmacology. It has the most terminology to memorise.
Stoichiometry (Duke Intro to Chemistry, Weeks 1–5)
The one method: everything goes through moles.
grams ──÷M──▶ MOLES ◀──×M── grams
particles ──÷6.022×10²³──▶ MOLES
gas litres (STP) ──÷22.4──▶ MOLES
solution: M × V(L) ──▶ MOLES
MOLES of A ──× (coef B / coef A)──▶ MOLES of B
| Quantity | Formula | Note |
|---|
| Moles | n = m / M | M = molar mass in g/mol |
| Avogadro | N = n × 6.022×10²³ | Atoms = molecules × atoms per molecule |
| Gas at STP | n = V / 22.4 L | 0 °C, 1 atm (IUPAC 1 bar uses 22.7 L) |
| Ideal gas | PV = nRT | R = 0.0821 L·atm/(mol·K) = 8.314 J/(mol·K). T in kelvin = °C + 273 |
| Molarity | M = n / V(L) | mol per litre of solution |
| Molality | m = n / kg solvent | Doesn't change with temperature |
| Dilution | M₁V₁ = M₂V₂ | Moles stay constant |
| Titration | n₁M₁V₁ = n₂M₂V₂ | n = H⁺ or OH⁻ count (H₂SO₄ = 2, NaOH = 1) |
| % composition | (mass of element in 1 mol / M) × 100 | Sum = 100% |
| % yield | actual / theoretical × 100 | Theoretical yield comes from the limiting reagent |
| Avg atomic mass | Σ (isotope mass × fractional abundance) | Cl: 0.75×35 + 0.25×37 = 35.5 |
Atomic masses: H 1 · C 12 · N 14 · O 16 · Na 23 · Mg 24 · Al 27 · S 32 · Cl 35.5 · K 39 · Ca 40 · Fe 56 · Cu 63.5 · Zn 65
Molar masses: H₂O 18 · CO₂ 44 · NaCl 58.5 · NaOH 40 · HCl 36.5 · H₂SO₄ 98 · CaCO₃ 100 · NH₃ 17 · CH₄ 16 · C₆H₁₂O₆ 180 · O₂ 32 · N₂ 28
Limiting reagent (most-tested)
- Convert each reactant to moles.
- Divide each by its coefficient in the balanced equation.
- Smallest value = limiting reagent. It sets the product amount. The others are in excess.
- Excess left = initial − amount used (via the mole ratio).
Example: 2H₂ + O₂ → 2H₂O with 4 g H₂ (2 mol) and 16 g O₂ (0.5 mol). Per coefficient: 2/2 = 1 and 0.5/1 = 0.5, so O₂ is limiting. H₂O = 1 mol = 18 g, and 1 mol of H₂ is left over.
Empirical and molecular formula
- Treat the % values as grams (assume a 100 g sample).
- Divide each by its atomic mass to get moles.
- Divide all by the smallest mole value.
- If you get ~.5, multiply by 2. If ~.33, multiply by 3.
- Molecular formula = (M ÷ empirical-formula mass) × empirical formula.
Example: C 40%, H 6.7%, O 53.3% → 3.33 : 6.7 : 3.33 → 1 : 2 : 1 → CH₂O (30). If M = 180, ×6 gives C₆H₁₂O₆.
Balancing equations
Balance metals first, then non-metals, then H, and O last. Keep polyatomic ions as units. Never change subscripts, only coefficients. CH₄ + 2O₂ → CO₂ + 2H₂O · C₃H₈ + 5O₂ → 3CO₂ + 4H₂O · 2Al + 3Cl₂ → 2AlCl₃ · N₂ + 3H₂ → 2NH₃.
Weeks 1–3 basics (measurement, atoms, bonding)
- Significant figures: non-zero digits count. Zeros between them count. Leading zeros never count (0.0045 → 2). Trailing zeros count only after a decimal point (2.50 → 3). ×/÷: the answer has the fewest sig figs of the inputs. +/−: the fewest decimal places.
- Accuracy = closeness to the true value. Precision = closeness of repeated readings to each other.
- Dimensional analysis: multiply by conversion factors so units cancel. 1 in = 2.54 cm. 1 L = 1000 mL = 1 dm³. 1 mL = 1 cm³.
- Density = mass/volume. Kelvin = °C + 273.15.
- Atom: protons (Z) set the element. Mass number A = protons + neutrons. Isotopes have the same Z and a different A. Cation = lost e⁻, anion = gained e⁻.
- Dalton's laws: conservation of mass, definite proportions, multiple proportions.
- Periodic trends: atomic radius ↓ across a period and ↑ down a group. Ionisation energy and electronegativity ↑ across and ↓ down.
- Bonding: ionic = metal + non-metal (electron transfer). Covalent = non-metals (electron sharing).
- Reaction types: combination, decomposition, single/double displacement, combustion, neutralisation, precipitation, redox (OIL RIG).
Common traps: using an unbalanced equation · forgetting the "2" in H₂/O₂/Cl₂ · mL instead of L in molarity · "atoms" vs "molecules" · picking the limiting reagent by grams instead of moles/coefficient · °C instead of K in PV=nRT.
Stoichiometry quiz
Chromatographic Separation (NPTEL Analytical Instrumentation)
Core idea: components separate because they distribute differently between a stationary phase and a mobile phase. A component that spends more time in the stationary phase elutes later. Invented by M. Tswett in 1903 to separate plant pigments.
| Classify by | Types |
|---|
| Mobile phase | Gas (GC), Liquid (LC/HPLC), Supercritical fluid (SFC, usually CO₂) |
| Mechanism | Adsorption (solid stationary phase) · Partition (liquid film) · Ion exchange (charged resin) · Size exclusion / gel permeation (pore size) · Affinity (specific bio-binding, e.g. antibody–antigen, His-tag–Ni) |
| Bed geometry | Column vs planar (TLC, paper) |
Formula box
| Term | Formula | Meaning |
|---|
| Retardation factor | Rf = d(solute) / d(solvent front) | TLC/paper; always 0–1 |
| Distribution constant | K = Cs / Cm | Higher K → longer retention |
| Adjusted retention time | t'R = tR − tM | tM = dead time (unretained peak) |
| Retention factor | k = (tR − tM) / tM | Ideal range is about 1–10 |
| Selectivity | α = k₂ / k₁ (k₂ > k₁) | α > 1 is needed for separation |
| Plate number | N = 16 (tR/W)² = 5.54 (tR/W½)² | W = base width, W½ = width at half height. Higher N = more efficient |
| Plate height (HETP) | H = L / N | Smaller H = more efficient |
| Resolution | Rs = 2(tR2 − tR1) / (W1 + W2) | Rs ≥ 1.5 = baseline separation. Rs = 1 → about 98% separated |
| Purnell equation | Rs = (√N/4)·((α−1)/α)·(k/(1+k)) | Rs ∝ √N ∝ √L. Doubling the length → Rs × 1.41 |
| Van Deemter | H = A + B/u + C·u | A = eddy diffusion (multiple paths; 0 in open-tubular columns) · B = longitudinal diffusion (dominates at low flow) · C = mass-transfer resistance (dominates at high flow). Optimum flow u = √(B/C) |
Gas chromatography (GC)
- Analytes must be volatile and thermally stable. Otherwise use derivatisation.
- Carrier gases: He, N₂, H₂. H₂ has the flattest Van Deemter curve, so it allows the fastest runs. N₂ gives the lowest minimum H, but only at slow flow.
- Columns: packed vs capillary/open-tubular (WCOT = wall-coated, SCOT = support-coated, PLOT = porous layer). Capillary columns give higher N.
- Temperature programming handles samples with a wide boiling range. It's the GC equivalent of HPLC gradient elution.
- Injector: split (concentrated samples) vs splitless (trace analysis).
- Kovats retention index uses n-alkanes as reference points (n-alkane = 100 × carbon number).
| GC detector | Key facts |
|---|
| FID (flame ionisation) | Most common. Responds to organic C–H compounds. Destructive. No response to H₂O, CO₂, N₂ or O₂. Wide linear range. |
| TCD (thermal conductivity) | Universal and non-destructive, but less sensitive. |
| ECD (electron capture) | Uses a ⁶³Ni β-source. Highly sensitive to halogens and other electronegative groups, e.g. pesticides and PCBs. |
| NPD / FPD | NPD: nitrogen and phosphorus. FPD (flame photometric): sulfur and phosphorus. |
| MS | Identifies compounds from mass spectra (GC-MS). |
HPLC
- Normal phase: polar stationary phase (silica) + non-polar mobile phase (hexane). Polar compounds elute last.
- Reverse phase (most common, ~75%+ of HPLC): non-polar stationary phase (C18/ODS, C8) + polar mobile phase (water/methanol/acetonitrile). Polar compounds elute first. Adding more organic solvent shortens retention.
- Isocratic = constant mobile-phase composition. Gradient = changing composition (better for complex samples).
- Components: solvent reservoir → degasser → pump → injector (loop/autosampler) → guard column → column → detector → data system.
- Particle size 3–5 µm. UHPLC uses < 2 µm: higher efficiency, but much higher back-pressure (up to ~1000+ bar).
- Peak area ∝ concentration. Quantify with a calibration curve or an internal standard.
- Peak tailing (common with basic compounds on silica) vs fronting (column overload). Tailing factor ≈ 1 is ideal.
| HPLC detector | Key facts |
|---|
| UV-Vis | Most common. Needs a chromophore. PDA/DAD captures the full spectrum and can check peak purity. |
| Fluorescence | Very sensitive and selective. Needs fluorescent analytes or derivatisation. |
| Refractive index (RI) | Universal but low sensitivity. Not gradient-compatible and temperature-sensitive. Used for sugars. |
| Electrochemical, ELSD, CAD, MS | ELSD/CAD suit non-chromophoric compounds. LC-MS gives identification. |
Other modes
- Size exclusion (SEC/GPC/gel filtration): large molecules elute first (they're excluded from the pores). Used for polymer MW distribution and protein aggregates.
- Ion exchange: cation exchanger has −ve groups (e.g. sulfonate –SO₃⁻) and binds cations. Anion exchanger has +ve groups (quaternary ammonium). Elute with a salt or pH gradient.
- Affinity: the most specific mode. Used for protein purification (Protein A for antibodies).
- TLC: silica plate. Less polar compounds travel farther (higher Rf) on silica.
Common traps: SEC order is reversed (big molecules first) · in reverse phase, polar compounds come out first · RI detectors can't run gradients · FID can't detect water · resolution scales with √N, not N · the B term matters at low flow, the C term at high flow.
Chromatography quiz
Statistical Inference (Johns Hopkins, Caffo)
Probability
- P(A∪B) = P(A) + P(B) − P(A∩B). Mutually exclusive → P(A∩B) = 0. Independent → P(A∩B) = P(A)P(B).
- Conditional: P(A|B) = P(A∩B)/P(B). Bayes: P(A|B) = P(B|A)P(A) / P(B).
- Diagnostic tests: Sensitivity = P(+|disease). Specificity = P(−|no disease). PPV = P(disease|+). NPV = P(no disease|−). Positive likelihood ratio = sens/(1−spec). Negative LR = (1−sens)/spec. PPV falls when prevalence is low.
- Trick: picture 1,000 people and fill a 2×2 table. It's faster than writing out Bayes.
Expectation and distributions
| Item | Formula / fact |
|---|
| Expected value | E[X] = Σ x·p(x). E[aX+b] = aE[X]+b |
| Variance | Var(X) = E[X²] − (E[X])². Var(aX+b) = a²Var(X). For independent X, Y: Var(X±Y) = Var X + Var Y |
| Standard error of the mean | SE = σ / √n. Quadrupling n halves the SE |
| Bernoulli | mean p, variance p(1−p) |
| Binomial | mean np, variance np(1−p). P(X=k) = C(n,k) p^k (1−p)^(n−k) |
| Poisson | mean = variance = λ. Models counts and rates |
| Normal | 68–95–99.7 rule. z: 1.645 (90%), 1.96 (95%), 2.576 (99%). Z = (X−μ)/σ |
| t distribution | Heavier tails than the normal. df = n−1. Approaches the normal as n grows. Use it when σ is unknown and n is small |
Asymptotics
- Law of Large Numbers: the sample mean converges to the population mean as n → ∞ (an estimator with this property is consistent).
- Central Limit Theorem: for large n, the sample mean is approximately N(μ, σ²/n), whatever shape the population has.
Confidence intervals
- Mean, σ known:
x̄ ± z·σ/√n. σ unknown: x̄ ± t(n−1)·s/√n.
- Proportion (Wald):
p̂ ± z√(p̂(1−p̂)/n). Quick 95% bound: p̂ ± 1/√n. Agresti–Coull adds 2 successes and 2 failures.
- Interpretation: if you repeated the procedure many times, about 95% of the intervals would contain the true parameter. It does not mean "95% probability that μ is in this particular interval".
- The interval gets wider with higher confidence or more variability, and narrower with larger n.
Hypothesis testing
| Concept | Key point |
|---|
| H₀ vs Hₐ | H₀ = no effect / status quo. Test statistic = (estimate − hypothesised value) / SE |
| Type I error (α) | Rejecting a true H₀ (false positive) |
| Type II error (β) | Failing to reject a false H₀ (false negative) |
| Power = 1 − β | Increases with larger n, larger effect size, larger α, and smaller variance. Usually 80–90% |
| p-value | The probability of data as extreme or more extreme than what was observed, assuming H₀ is true. Reject if p < α. It is NOT P(H₀ is true) |
| CI ↔ test | A 95% CI excluding the H₀ value ⇔ reject at α = 0.05 (two-sided) |
| Paired vs independent | Same subjects before/after → paired t-test on the differences. Two separate groups → two-sample t (pooled if variances are equal, otherwise Welch) |
| Multiple testing | Bonferroni: α/m (controls the family-wise error rate). Benjamini–Hochberg: controls the false discovery rate (FDR) and is less strict |
| Resampling | Bootstrap: resample with replacement from the data to estimate a sampling distribution or CI. Permutation test: shuffle group labels to build the null distribution |
Common traps: p-value ≠ the probability that H₀ is true · "fail to reject" ≠ "accept" · SE ≠ SD · Var(X−Y) adds the variances · CIs shrink with √n, not n · statistical significance ≠ practical importance.
Statistical Inference quiz
Drug Development (UCSD)
The big numbers: about 10–15 years from discovery to approval, roughly $1–2.6 billion per approved drug (Tufts estimate $2.6B), about 1 approval per 5,000–10,000 compounds screened, and about 10% of drugs entering Phase I get approved. Phase II has the highest attrition, mostly from lack of efficacy.
Pipeline
| Stage | Who / how many | Purpose |
|---|
| Preclinical | In vitro + animals (2 species: rodent + non-rodent), under GLP | Pharmacology, PK, toxicology. Find the NOAEL (no observed adverse effect level) to set the first human dose |
| IND filing | Submitted to FDA | Permission to start human trials. FDA has 30 days to object (clinical hold) |
| Phase 0 | ~10–15, microdose | Exploratory PK (optional) |
| Phase I | 20–100 healthy volunteers (cancer drugs: patients) | Safety, tolerability, PK, maximum tolerated dose (MTD) |
| Phase II | 100–several hundred patients | Early efficacy, dose-finding, short-term side effects |
| Phase III | 300–3,000+ patients, often multicentre | Confirm efficacy vs placebo or standard care. These are the pivotal trials used for approval |
| NDA / BLA | To FDA: CDER (drugs) / CBER (many biologics) | Marketing approval. Standard review ~10 months, priority review ~6 months (PDUFA goals) |
| Phase IV | Real-world population | Post-marketing surveillance and pharmacovigilance; rare and long-term effects |
Regulatory terms
GxP
GLP = Good Laboratory Practice (nonclinical studies) · GMP/cGMP = Good Manufacturing Practice · GCP = Good Clinical Practice (ICH E6) · GDP = Good Distribution / Documentation Practice.
Expedited pathways
Fast Track (serious condition, unmet need) · Breakthrough Therapy (substantial improvement) · Accelerated Approval (based on a surrogate endpoint; confirmatory trial required) · Priority Review (6 months).
Generics
ANDA (Abbreviated NDA, Hatch-Waxman Act 1984). Must show bioequivalence: the 90% CI of the AUC and Cmax ratio must fall within 80–125%. No repeat efficacy trials. Biologic copies are biosimilars.
Orphan drugs
US: disease affecting < 200,000 people (Orphan Drug Act 1983). 7 years market exclusivity in the US, 10 in the EU.
Regulators
US: FDA · EU: EMA (MAA) · Japan: PMDA · UK: MHRA · India: CDSCO, headed by the DCGI · Harmonised guidelines from ICH.
History
Thalidomide (birth defects, about 1960) → Kefauver–Harris Amendment 1962, which required proof of efficacy. 1938 FD&C Act (safety) came after the sulfanilamide elixir deaths.
CMC
Chemistry, Manufacturing and Controls: API synthesis, formulation (API + excipients), specifications, stability, scale-up, process validation.
Data and records (Zifo-relevant)
21 CFR Part 11 = FDA rules for electronic records and signatures. ALCOA+ data integrity. Audit trails. Computer system validation (CSV).
Human PK basics
- Cmax = peak plasma concentration. Tmax = time to reach it. AUC = total exposure. t½ = half-life. CL = clearance. Vd = volume of distribution.
- Bioavailability F = AUC(oral)/AUC(IV). IV = 100%.
- Steady state is reached after about 4–5 half-lives.
- Patent term is 20 years from filing, so much of it is used up during development. Patent term extension partly makes up for this.
- Safety oversight: informed consent, IRB / Ethics Committee, DSMB, adverse event reporting (FAERS), REMS for high-risk drugs.
Common traps: IND (before humans) vs NDA (before marketing) · Phase I usually uses healthy volunteers · Phase II (not III) has the highest failure rate · bioequivalence is 80–125%, not ±20% · BLA is for biologics · orphan threshold is 200,000 in the US.
Drug Development quiz
Design and Interpretation of Clinical Trials (Johns Hopkins)
The randomised controlled trial (RCT) is the gold standard because randomisation balances both known and unknown confounders between groups.
Design types
| Design | What / when |
|---|
| Parallel | Each participant gets one treatment for the whole trial. The most common design |
| Crossover | Each participant gets both treatments in sequence, with a washout period between. Acts as its own control, so it needs fewer subjects. Only for chronic, stable conditions and short-acting, reversible treatments. Risk: carry-over effect |
| Factorial (2×2) | Tests two interventions at once (A, B, A+B, neither). Efficient if they don't interact |
| Cluster randomised | Randomises groups (schools, villages, hospitals). Used when contamination between individuals is likely. Needs a larger sample (intracluster correlation) |
| Superiority / Equivalence / Non-inferiority | Non-inferiority: show the new treatment is not worse than the active control by more than a prespecified margin (e.g. it's cheaper or safer) |
| Adaptive | Prespecified changes based on interim data (dose arms dropped, sample size re-estimated) |
Randomisation and masking
- Simple randomisation: like a coin flip. Group sizes can end up unequal in small trials.
- Block randomisation: keeps group sizes balanced over time (e.g. blocks of 4: AABB, ABAB…).
- Stratified randomisation: randomises separately within strata of a key prognostic factor (site, disease stage) so groups are balanced on it.
- Allocation concealment: hides the upcoming assignment from recruiters, preventing selection bias. It happens before assignment and is always possible.
- Masking/blinding: hides the assignment after randomisation (single = participant, double = participant + investigator/assessor, triple = + data analyst). Prevents performance and ascertainment (detection) bias and placebo effects. It isn't always possible (e.g. surgery).
Outcomes
- Primary outcome: one, prespecified. It drives the sample size. Secondary outcomes are supportive.
- Clinical endpoint: how a patient feels, functions or survives (death, MI, stroke).
- Surrogate endpoint: a lab or physical measure that stands in for the clinical outcome (blood pressure, LDL cholesterol, HbA1c, CD4 count, tumour size). Faster, but it can mislead.
- Composite endpoint: several events combined (e.g. MI + stroke + CV death) to increase event rates.
Sample size and analysis
- Sample size depends on α (usually 0.05), power (80–90%), effect size and variability. A bigger effect or less variability means a smaller n. Higher power or a smaller α means a larger n. Inflate for expected dropouts.
- Intention-to-treat (ITT): analyse everyone in the group they were randomised to, regardless of adherence. It preserves randomisation, is conservative, and is the primary analysis for superiority trials.
- Per-protocol: only participants who adhered. Prone to bias, but important in non-inferiority trials alongside ITT.
- Interim analysis by an independent DSMB (Data Safety Monitoring Board) can stop a trial early for benefit, harm or futility (e.g. O'Brien–Fleming boundaries).
- Subgroup analyses should be prespecified and interpreted with caution (multiple-testing risk).
Effect measures
| Measure | Formula (control risk CER, treated risk EER) |
|---|
| Absolute risk reduction | ARR = CER − EER |
| Relative risk | RR = EER / CER |
| Relative risk reduction | RRR = ARR / CER = 1 − RR |
| Number needed to treat | NNT = 1 / ARR (round up) |
| Odds ratio / Hazard ratio | OR for case-control or logistic models. HR for time-to-event (survival). A value of 1 = no difference |
Example: CER 20%, EER 15% → ARR 5% → NNT 20. RR 0.75 → RRR 25%.
Ethics and reporting
- Nuremberg Code (1947, voluntary consent) → Declaration of Helsinki (1964, WMA) → Belmont Report (1979): respect for persons, beneficence, justice. Tuskegee syphilis study was the key abuse case.
- Equipoise: genuine uncertainty about which arm is better. It's the ethical basis for randomising.
- Informed consent + IRB/Ethics Committee approval. Trials are registered (ClinicalTrials.gov, CTRI in India).
- CONSORT = reporting guideline for RCTs (checklist + participant flow diagram).
- Bias types: selection, performance, detection/ascertainment, attrition (loss to follow-up), reporting.
Common traps: allocation concealment ≠ blinding · ITT analyses people as randomised, not as treated · crossover needs a washout and a stable disease · a surrogate is not a clinical outcome · NNT = 1/ARR, not 1/RRR.
Clinical Trials quiz
Drug Discovery (UCSD Skaggs School)
Target ID → Target validation → Assay development → Hit ID (HTS / virtual / fragment) → Hit-to-lead → Lead optimisation → Preclinical candidate
Targets and sources
- Target classes: GPCRs (largest share, ~30–35% of approved drugs), enzymes (kinases, proteases), ion channels, nuclear receptors, transporters.
- Target validation: show that modulating the target changes the disease (knockouts, siRNA/CRISPR, genetic association, tool compounds).
- Sources of drugs: natural products (aspirin from willow bark salicin, penicillin from Penicillium (Fleming 1928), taxol/paclitaxel from the Pacific yew, artemisinin from sweet wormwood, morphine from the poppy), synthetic libraries, biologics, repurposing.
- Small molecules (< ~900 Da, often oral) vs biologics (antibodies "-mab", proteins; large, injected, highly specific).
Screening and optimisation
- HTS: robotic screening of 10⁵–10⁶ compounds in 96/384/1536-well plates. Assay quality metric Z′-factor: 0.5–1 = excellent.
- Virtual screening / structure-based design (docking against an X-ray or cryo-EM structure). Fragment-based discovery (small fragments, < 300 Da, grown into leads).
- SAR (structure–activity relationship): how changing the structure changes activity. Pharmacophore = the essential 3D features needed for binding. Bioisostere = a group swap with similar properties.
- Lead optimisation improves potency, selectivity, and ADMET properties.
- Lipinski's Rule of 5 (oral drug-likeness): MW ≤ 500, logP ≤ 5, H-bond donors ≤ 5, H-bond acceptors ≤ 10. More than one violation → likely poor oral absorption.
Pharmacodynamics (what the drug does to the body)
| Term | Meaning |
|---|
| IC50 / EC50 | Concentration giving 50% inhibition / 50% of maximum effect. Lower = more potent |
| Kd / Ki | Dissociation / inhibition constant. Lower = tighter binding (higher affinity) |
| Potency vs efficacy | Potency = how much drug is needed. Efficacy (Emax) = maximum achievable effect |
| Agonist | Binds and activates (full agonist = max effect; partial agonist = submaximal) |
| Antagonist | Binds without activating. Competitive: dose–response curve shifts right in parallel, Emax unchanged (surmountable). Non-competitive: Emax decreases |
| Inverse agonist | Reduces constitutive (baseline) receptor activity |
| Therapeutic index | TI = TD50 / ED50 (or LD50/ED50). Larger = safer |
| Enzyme inhibition | Competitive: Km ↑, Vmax unchanged · Non-competitive: Vmax ↓, Km unchanged · Uncompetitive: both ↓ · Irreversible: covalent (aspirin acetylates COX) |
Pharmacokinetics: ADME (what the body does to the drug)
- Absorption → Distribution → Metabolism → Excretion (+ Toxicity = ADMET).
- Bioavailability (F): the fraction reaching systemic circulation. IV = 100%. Oral drugs lose some to first-pass metabolism in the liver.
- Vd = dose / C₀. Half-life
t½ = 0.693 × Vd / CL. Steady state after 4–5 half-lives.
- Phase I metabolism: oxidation, reduction, hydrolysis, mainly by cytochrome P450. CYP3A4 metabolises about 50% of drugs. Inhibitors and inducers cause drug–drug interactions (grapefruit juice inhibits CYP3A4).
- Phase II metabolism: conjugation (glucuronidation, sulfation, acetylation, glutathione). Makes the drug more water-soluble so it can be excreted, mainly by the kidneys.
- Prodrug: inactive until metabolised (enalapril → enalaprilat, codeine → morphine, levodopa → dopamine).
- Key toxicity flag: hERG K⁺ channel blockade → QT prolongation / cardiac arrhythmia. Also hepatotoxicity and genotoxicity (Ames test).
Common traps: lower IC50 = more potent (not less) · a competitive antagonist doesn't lower Emax · competitive enzyme inhibition raises Km, not Vmax · Phase I/II metabolism ≠ Phase I/II trials · the Rule of 5 is about multiples of 5 (MW 500, HBA 10).
Drug Discovery quiz
Essay framework (15 min)
Candidates report a picture-based essay. This template works for both a picture and a topic prompt.
0–2 min · Plan
Note 3 things you see, one theme they suggest, and 2 arguments plus 1 example. Choose a one-line thesis.
2–13 min · Write
4–5 paragraphs, 250–350 words. Short, clear sentences. About 25 words per minute.
13–15 min · Review
Fix spelling, tense and subject–verb errors. Check that your thesis appears in both the intro and the conclusion.
Template: O-I-C-C (Observe → Interpret → Connect → Conclude)
- Intro (Observe + thesis): "The image shows ___, ___ and ___. At first glance it depicts ___, but it really reflects ___." End with your thesis.
- Body 1 (Interpret): What does the picture symbolise? The main idea, plus one reason why it matters.
- Body 2 (Connect, real world): Link it to science, technology, R&D, society or the workplace. Give one concrete example (AI in drug discovery, lab automation, COVID-19 vaccine timelines, data integrity).
- Body 3 (Balance): One challenge or counterpoint, and how it can be addressed. This shows analytical maturity.
- Conclusion: Restate the thesis in new words, add a forward-looking line, and connect it to yourself ("As a future professional, I…").
What graders look for: staying on topic (relevance to the picture), clear structure, grammar, vocabulary range, and a point of view. Don't just describe the image; interpret it.
Reading common picture types
| If the picture shows… | Angle to take |
|---|
| A team rowing, climbing or building together | Collaboration, cross-functional teams (scientists + IT), shared goals |
| A lab, microscope, DNA helix or test tubes | Science accelerated by data and informatics, digital labs, reproducibility |
| A person at a crossroads, or on a ladder or stairs | Decision-making under uncertainty, growth mindset, continuous learning |
| Robots, AI, screens, a human hand touching a robot hand | Automation augmenting humans, ethics, upskilling |
| Nature vs a factory, a melting globe, a plant in hands | Sustainability, green chemistry, responsible innovation |
| A puzzle, a maze, a lightbulb, a chess board | Problem-solving, innovation, strategy, first principles |
| A crowded city or a person alone with a phone | Digital connectivity vs human connection, mental health, balance |
| One different-coloured fish swimming against the school | Thinking differently, courage, leadership, innovation |
| A mountain summit or flag, a runner at the finish line | Perseverance, goal-setting, learning from failure |
Practice prompts
- How is digital transformation changing pharmaceutical R&D?
- AI in drug discovery: hype or the next breakthrough?
- Why data integrity (ALCOA+) matters in scientific research
- Bridging science and technology: the role of scientific informatics
- Lab automation and ELN/LIMS: the lab of the future
- Lessons from the speed of COVID-19 vaccine development
- Is failure necessary for innovation?
- Remote work vs collaboration in research teams
- Ethics of using patient data in research
- Technology: a boon or a bane for society?
- The role of youth in building a sustainable future
Useful vocabulary: FAIR data (Findable, Accessible, Interoperable, Reusable) · ALCOA+ (Attributable, Legible, Contemporaneous, Original, Accurate + Complete, Consistent, Enduring, Available) · LIMS · ELN · lab digitalisation · cloud informatics · regulatory compliance (GxP, 21 CFR Part 11) · time-to-market · reproducibility · interdisciplinary · paradigm shift · catalyst · synergy.
Full sample essay: picture of a team building a bridge (~300 words)
The image shows a group of people placing planks to build a bridge across a deep gap. Each person holds a single piece, and none of them could cross alone. On the surface it is about construction, but it captures how modern progress happens: breakthroughs are built together, one contribution at a time.
The bridge represents a shared goal, and each plank represents an individual's skill. The structure is only as strong as its weakest link, so trust and coordination matter as much as talent. The picture also suggests that the gap, whether a problem, a disease or a market need, cannot be crossed by the most gifted individual working in isolation.
Scientific research today is a clear example. Developing a new medicine needs biologists, chemists, clinicians, statisticians and software engineers. During the COVID-19 pandemic, vaccines were developed in under a year because data, labs and regulators worked in parallel. Scientific informatics is the bridge in such work, connecting instruments, data and people so that discoveries are faster and reproducible.
Collaboration has challenges. Teams from different disciplines speak different "languages", and poor communication can cause delays or errors. These can be overcome with common data standards, clear roles and a culture where people ask questions openly.
Like the bridge in the image, the future of research will be built plank by plank, by scientists, engineers and data experts working as one. As someone entering the field of scientific informatics, I hope to be one of those planks: connecting science with technology so that discoveries reach people faster.
Sample openings for 4 more pictures
- Robot and human hands touching: "The image of a human hand reaching toward a robotic one captures the defining relationship of our era: not humans versus machines, but humans with machines…"
- Person at a crossroads: "A lone figure stands where two roads split, one paved and one overgrown. Every career and every experiment begins at such a junction, where the safe choice and the bold one diverge…"
- Lightbulb made of puzzle pieces: "The lightbulb, the universal symbol of an idea, is shown here assembled from puzzle pieces. Innovation rarely arrives as a single flash; it is assembled…"
- Plant growing from coins: "A small sapling rising out of a stack of coins suggests that growth, whether financial, personal or scientific, needs patient investment…"
Gamified assessment (20 min)
These games test how you think, not what you know. They usually measure accuracy and speed, and some also track consistency and how you handle risk.
| Game type | What it looks like | How to approach |
|---|
| Ball into hole / route puzzle reported | Tilt, rotate or place pipes/blocks so a ball reaches the target. Fewer moves = higher score. | Trace the path backward from the hole before your first move. Count the minimum moves. Don't trial-and-error fast; extra moves cost points. |
| Multitasking reported | Two tasks at once (e.g. sort shapes left while answering colours right). | Accuracy first, then speed. Use a steady rhythm and don't panic after one mistake. Glance between the tasks on a fixed cadence. |
| Coded diagrams / symbol logic reported | Shapes go through "machines" that transform them. Predict the output or work out the rule. | Work out what each operator does from the first 1–2 examples (rotate, colour swap, flip, add/remove), then apply them in order. Write the rules down. |
| Multi-level puzzles reported | Grid or block puzzles (sliding, Tower of Hanoi, switch lights) that get harder each level. | Plan 2–3 moves ahead. In Tower of Hanoi the minimum is 2ⁿ−1 moves, so move the smallest disk every other turn. |
| Memory span | A sequence of tiles, digits or colours flashes. Repeat it, sometimes in reverse. | Chunk into groups of 3. Say them under your breath. Use spatial patterns ("L-shape, then corner"). |
| Pattern / grid reasoning | Find the missing tile, or the odd one out. | Check rows, then columns, then diagonals. Look at count, rotation, shading and position. |
| Risk–reward (balloon pump) | Each pump earns money, but the balloon can burst and you lose that round. | Be moderate and consistent (bank at about the midpoint). Extreme caution or recklessness both read as poor judgment. |
| Attention / Stroop / n-back | Respond to a word's ink colour rather than the word, or check whether this item matches the one 2 back. | Keep your eyes on the centre. Answer the actual rule, not your reflex. Go slightly slower for accuracy. |
Mindset: the first level is usually a tutorial, so read the rules fully. Difficulty adapts: if levels get harder, you're doing well. A bad level resets nothing, so take one breath and treat the next one fresh. Use a mouse if you can, and close every other app.
Practice tonight (10–15 min): search for "Pymetrics-style games", "BART balloon task online", "n-back game online", "Tower of Hanoi online" and "pipe connect puzzle", and play a few rounds of each to warm up.
Interview prep (next round)
Candidates report self-intro, resume and project questions, puzzles solved aloud, and HR questions.
Know the company: Zifo RnD Solutions is a Chennai-headquartered (founded 2008) scientific informatics services company. It works for pharma, biotech, chemicals, FMCG and other R&D-driven organisations worldwide. It implements and supports lab informatics platforms (LIMS, ELN, SDMS, CDS), scientific data management, lab automation and instrument integration, data science and analytics, and GxP/compliance services. The company is positioned as "science + IT", so they hire science and engineering graduates and train them.
| Likely question | How to answer |
|---|
| Tell me about yourself | 60 s: education → one strong project → skills (science + tech) → why this role. |
| Why Zifo? | It sits where science and technology meet; I can use my science background while building tech skills; a structured training programme; global pharma clients; real impact on how fast medicines are developed. |
| What is LIMS / ELN? | LIMS = Laboratory Information Management System (samples, tests, results, workflows). ELN = Electronic Lab Notebook (replaces paper notebooks for experiments). Both support data integrity and compliance. |
| Explain your project | STAR: Situation, Task, Action (what you did), Result (numbers if possible). Be ready for "what would you improve?" |
| Are you OK with the 2-year service agreement / relocation / shifts? | Answer clearly and positively: "Yes. I'm looking for long-term growth, and the training is valuable." |
| Strengths / weaknesses | A strength with an example. A real but non-fatal weakness, plus what you're doing about it. |
| Where do you see yourself in 5 years? | Growing into a domain expert / consultant in lab informatics, leading client projects. |
| Puzzles aloud | Think out loud, state assumptions, try small cases. Classics: 3 bulbs and 3 switches; 2 eggs and 100 floors; 8 balls with 1 heavier (2 weighings); river-crossing problems; 3- and 5-litre jugs to measure 4 L. |