Genomic medicine: GWAS, polygenic risk scores, gene therapy (CRISPR-Cas9 therapeutic applications)

Intuition Beginner

The Human Genome Project (1990–2003) read all 3 billion DNA letters for $3billion;todaysequencingcostsabout$200. Genome-wide association studies (GWAS) compare DNA of thousands of sick and healthy people, finding disease-linked variants — over 500,000 catalogued. Polygenic risk scores combine hundreds of variants to predict individual risk for heart disease, breast cancer, and diabetes. Gene therapy was a dream until 2012, when Jennifer Doudna and Emmanuelle Charpentier described CRISPR-Cas9 — a bacterial immune mechanism repurposed as precision gene-editing scissors. CRISPR corrects disease-causing mutations: sickle cell disease was cured in trials, and Casgevy became the first FDA-approved CRISPR therapy in 2023. But CRISPR raises ethical alarms — He Jiankui created gene-edited babies in 2018, sparking global condemnation.

Visual Beginner

Concept	What it does	Key fact
GWAS	Compares SNP frequencies in cases vs. controls	500,000+ associations found
Polygenic risk score	Combines hundreds of variant effects	Top 5% risk ~ 3x population average
CRISPR-Cas9	Cuts DNA at a programmed 20-nt target	Casgevy (2023): first FDA-approved CRISPR therapy
Base editing	Changes a single nucleotide without a break	C to T, A to G (Liu)
Prime editing	Search-and-replace genome editing	Liu 2019; no double-strand break

Gene therapy	Disease	Delivery	Year
Luxturna	Inherited retinal dystrophy	AAV	2017
Zolgensma	Spinal muscular atrophy	AAV9	2019
Casgevy	Sickle cell disease	Ex vivo CRISPR in HSCs	2023

Worked example Beginner

Worked example: a polygenic risk score for coronary artery disease

A 45-year-old man with no family history undergoes clinical PRS testing. His score places him in the 98th percentile for coronary artery disease risk.

Step 1. Khera et al. (2018) showed that individuals in the top 5% of the PRS distribution carry roughly 3-fold higher risk than the population average — comparable to monogenic Familial Hypercholesterolemia ^{[Nature Genetics 50 (2018) 1219-1224]}. For this patient, the top-2% score implies a lifetime risk approaching 30%, versus ~8% population baseline (see 35.03.02, cardiovascular disease).

Step 2. The clinical action: initiate statin therapy earlier than guidelines based on age alone would suggest, target a more aggressive LDL level (< 70 mg/dL), and intensify lifestyle counselling. The PRS does not diagnose disease — it stratifies risk to guide prevention.

Step 3. The limitation: this PRS was trained on European-ancestry GWAS data. Performance in other ancestries degrades because effect sizes, allele frequencies, and linkage patterns differ across populations (see 35.06.02, health disparities). Clinical deployment requires ancestry-specific calibration.

Check your understanding Beginner

Question 1: Why is the GWAS significance threshold p < 5 × 10⁻⁸ rather than p < 0.05?

A) Larger sample sizes demand stricter thresholds B) Bonferroni correction for ~10⁶ independent SNP tests C) It is arbitrary convention D) Population stratification requires it

Answer: B. Testing a million SNPs at p < 0.05 would yield ~50,000 false positives by chance. The Bonferroni-corrected threshold 0.05 / 10⁶ = 5 × 10⁻⁸ controls the family-wise false-positive rate.

Question 2: Casgevy treats sickle cell disease by editing which target?

A) The mutated β-globin gene directly B) The BCL11A enhancer to reactivate fetal hemoglobin C) The Cas9 gene itself D) An AAV vector delivered in vivo

Answer: B. Casgevy edits the BCL11A enhancer in hematopoietic stem cells ex vivo, reactivating fetal hemoglobin (HbF), which compensates for defective adult hemoglobin.

Question 3: True or false: a PRS trained on European-ancestry GWAS data performs equally well in all populations.

Answer: False. Approximately 78% of GWAS participants are of European ancestry. PRS trained on these data transfer poorly to other ancestries, risking widened health disparities.

Formal definition Intermediate+

Genome-wide association studies (GWAS). A GWAS tests each of 500,000+ single nucleotide polymorphisms (SNPs) for association with case-control status via logistic regression ^{[Ch. 10 Gene Mapping and GWAS]}. Significance is set at p < 5 × 10⁻⁸ — a Bonferroni correction for ~10⁶ independent tests (see 29.01.03). Individual odds ratios are small (1.1–1.5), and their aggregate partially explains the "missing heritability" gap (see 19.05.03, quantitative genetics). Population stratification — ancestry differences between cases and controls — is corrected via principal-component covariates (see 31.04.03).

Polygenic risk scores (PRS). A PRS combines effects across n variants:

$$\text{PRS}={\beta }_1{G}_1+{\beta }_2{G}_2+\cdots +{\beta }_n{G}_n$$

where βᵢ is the GWAS-estimated effect size and Gᵢ ∈ {0, 1, 2} is the allele count. Khera et al. showed the top 5% of PRS for coronary artery disease carry risk comparable to monogenic Familial Hypercholesterolemia ^{[Nature Genetics 50 (2018) 1219-1224]}. Transferability across ancestries is limited by Eurocentric discovery cohorts (see 35.06.02).

CRISPR-Cas9. Cas9 is a nuclease guided to a 20-nt target by a single guide RNA (sgRNA). Upon PAM recognition (5′-NGG-3′), Cas9 creates a double-strand break ^{[Science 337 (2012) 816-821]}. Repair via non-homologous end joining (NHEJ) produces gene knockouts; homology-directed repair (HDR) introduces precise edits from a donor template. Base editing and prime editing (Liu 2019) achieve nucleotide changes without double-strand breaks. Casgevy (2023) edits the BCL11A enhancer ex vivo to reactivate fetal hemoglobin for sickle cell disease (see 33.06.*, genetic engineering; 35.03.04).

Key result Intermediate+

The GWAS significance threshold via the Bonferroni correction

The nontrivial question is why the genome-wide significance threshold is p < 5 × 10⁻⁸ rather than the conventional p < 0.05 used elsewhere in statistics.

Setup. A GWAS tests m ≈ 10⁶ independent common SNPs for association. Testing each at α = 0.05 would yield m × α = 50,000 expected false positives — an uninterpretable result.

Claim. If m true null hypotheses are each tested at level α / m, the family-wise error rate (FWER — probability of at least one false positive) is bounded by α.

Proof. Let Fᵢ denote the event that test i yields a false positive. Under each null, P(Fᵢ) = α / m. The FWER is:

$$\text{FWER}=P(F_1\cup F_2\cup \cdots \cup F_m).$$

By the union bound (Boole's inequality):

$$P(F_1\cup \cdots \cup F_m)\le P(F_1)+P(F_2)+\cdots +P(F_m)=m\cdot \frac{\alpha }{m}=\alpha .$$

Application. With m = 10⁶ and α = 0.05, the per-test threshold is 0.05 / 10⁶ = 5 × 10⁻⁸. The expected false-positive count drops from 50,000 to 0.05 — fewer than one per GWAS. This threshold is the field standard drawn as a dashed line on every Manhattan plot (see 29.01.03, statistical reasoning, multiple testing).

Exercises Intermediate+

Exercise 1 (easy). A GWAS identifies a variant with odds ratio 1.3 and p = 3 × 10⁻⁹ for type 2 diabetes. Is this genome-wide significant? Why might the effect size be small even if the association is real?

Exercise 2 (medium). Explain why testing 10⁶ SNPs at α = 0.05 without correction produces ~50,000 false positives. Derive the Bonferroni-corrected threshold and state the condition on which the union bound relies.

Exercise 3 (medium). Contrast NHEJ and HDR as DNA repair outcomes following Cas9 cleavage. Which is preferred for creating a gene knockout, and which for precise correction of a point mutation?

Exercise 4 (hard). Distinguish somatic from germline gene editing. Explain why the scientific community condemned He Jiankui's 2018 CCR5 edits in twin embryos despite the technical feasibility of the procedure.

Exercise 5 (hard). A PRS for breast cancer achieves AUC 0.65 in a European-ancestry cohort but drops to 0.55 in an African-ancestry cohort. Propose three reasons for the degradation and discuss the ethical implications of clinical deployment under these conditions.

Advanced results Master

Genomic medicine and health equity

Roughly 78% of GWAS participants are of European ancestry (Martin et al. 2019), limiting PRS transferability: scores trained on European data lose 30–60% of predictive power in African-ancestry populations. The H3Africa consortium and the NIH All of Us program address this gap. Population-specific founder mutations — BRCA in Ashkenazi Jewish populations, the Finnish Disease Heritage — illustrate that allele spectra differ by ancestry (see 35.06.02, health disparities; 31.06.*, anthropology). Race is non-biological, but genetic ancestry is real and clinically relevant (see 31.04.03, race as non-biological).

Gene editing ethics

Somatic editing (affecting only the treated individual) is ethically comparable to other interventions once safety is established. Germline editing — heritable changes in embryos — is far more controversial. He Jiankui's 2018 CCR5 edits in twin girls violated scientific consensus, carried unknown off-target risks, and resulted in a prison sentence. The WHO subsequently issued governance recommendations (see 20.02., bioethics; 20.07., democratic governance). The therapy-enhancement boundary — treating disease versus augmenting traits — remains contested, and the shadow of eugenics history demands vigilance (see 19.05., eugenics; 29.05., cognition).

Pan-omics and the future

Integration of genomic, transcriptomic, proteomic, and metabolomic data is moving medicine toward multi-omic precision. The T2T consortium completed the first gapless human genome in 2022. Epigenomic profiling adds regulatory context (see 17.06.04, epigenetics). Single-cell genomics reveals tumor heterogeneity (see 35.03.03). Pathogen genomics underpins surveillance — COVID variant tracking and vaccine design (see 35.02.03; 35.06.03).

Connections Master

Quantitative genetics

GWAS and PRS operationalize the quantitative-genetics framework: complex traits arise from many variants of small effect, each consistent with the polygenic model. Missing heritability — the gap between GWAS-explained variance and twin-study estimates — remains an active question (19.05.03, quantitative genetics; 19.05.*, polygenic adaptation).

Statistical reasoning

GWAS methodology — multiple-testing correction, Bonferroni and FDR thresholds, population stratification correction — applies statistical inference to high-dimensional genetic data (29.01.03, statistical reasoning; 37.*, probability and statistics).

Pharmacology and pharmacogenomics

Genomic variation guides drug response: CYP2D6 polymorphisms alter codeine metabolism; HLA-B*57:01 predicts abacavir hypersensitivity. Pharmacogenomic testing personalizes prescribing (35.07.02, pharmacokinetics; 35.07.03, drug classes; 31.04.03, genetic variation in drug response).

Cancer biology

Tumor sequencing (TCGA), mutational signatures, circulating tumor DNA (ctDNA) for liquid biopsy, and tumor mutational burden as an immunotherapy biomarker all apply genomic tools to oncology (35.03.03, cancer biology, genomics and liquid biopsies).

Health disparities and ethics

Eurocentric bias in genomic databases, $2.1M pricing for Zolgensma, and germline-editing governance place genomic medicine within the political economy of health (35.06.02, health disparities; 30.04., sociology, access to care; 20.02., ethics, bioethics and genetic privacy; 36.*, media literacy, data privacy).

Downstream link

Genomic medicine provides the biological substrate — variant catalogs, risk models, gene-editing tools — upon which AI-driven precision medicine builds for clinical decision support and drug response prediction (35.08.03, precision medicine and AI).

Historical and philosophical context Master

The Human Genome Project (1990–2003)

The HGP produced the first draft human genome in 2001 ^{[Nature 409 (2001) 860-921]}. Expectations that the genome would quickly reveal most disease genes proved optimistic: common diseases are polygenic, involving hundreds of small-effect variants interacting with environment. Only ~1.5% of the genome codes for proteins; the remainder encodes regulatory elements. The finding that humans carry ~20,000 protein-coding genes — fewer than rice — humbled gene-centric biology (see 17.06.*).

CRISPR: from bacterial immunity to therapy (1987–2023)

CRISPR loci were first described in E. coli in 1987 as mysterious repetitive sequences. Two decades of microbiology revealed them as a bacterial adaptive immune system: bacteria archive viral DNA fragments as spacers, transcribe them into guide RNAs that direct nucleases to destroy matching invaders. Doudna and Charpentier's 2012 Science paper showed Cas9 could be reprogrammed with a chimeric guide RNA to cut any DNA target ^{[Science 337 (2012) 816-821]}, earning the 2020 Nobel Prize in Chemistry. The arc from basic microbiology to therapy (Casgevy, 2023) exemplifies how curiosity-driven science drives medicine.

The eugenics shadow

Genomic medicine inherits the legacy of eugenics — the early-20th-century movement that invoked genetics to justify forced sterilization. Distinguishing legitimate genetic medicine from eugenic selection requires vigilance as germline editing and polygenic embryo selection become feasible. The He Jiankui affair showed that ethical infrastructure can lag behind technical capability, and that individual scientists can act unilaterally on decisions with heritable consequences (see 20.02., ethics; 35.06., eugenics history).

Bibliography Master

1. International Human Genome Sequencing Consortium. (2001). "Initial Sequencing and Analysis of the Human Genome." Nature, 409(6822), 860–921.

2. Jinek, M., Chylinski, K., Fonfara, I., Hauer, M., Doudna, J. A. & Charpentier, E. (2012). "A Programmable Dual-RNA-Guided DNA Endonuclease in Adaptive Bacterial Immunity." Science, 337(6096), 816–821.

3. Cong, L., Ran, F. A., Cox, D. et al. (2013). "Multiplex Genome Engineering Using CRISPR/Cas Systems." Science, 339(6121), 819–823.

4. Khera, A. V., Chaffin, M., Aragam, K. G. et al. (2018). "Genome-wide Polygenic Scores for Common Diseases Identify Individuals with Risk Equivalent to Monogenic Mutations." Nature Genetics, 50(9), 1219–1224.

5. Martin, A. R., Kanai, M., Kamatani, Y. et al. (2019). "Clinical Use of Current Polygenic Risk Scores May Exacerbate Health Disparities." Nature Genetics, 51(4), 584–591.

6. Frangoul, H., Altshuler, D., Cappellini, M. D. et al. (2021). "CRISPR-Cas9 Gene Editing for Sickle Cell Disease and β-Thalassemia." New England Journal of Medicine, 384(3), 252–260.

7. Anzalone, A. V., Randolph, P. B., Davis, J. R. et al. (2019). "Search-and-Replace Genome Editing Without Double-Strand Breaks or Donor DNA." Nature, 576(7785), 149–157.

8. Strachan, T. & Read, A. P. (2019). Human Molecular Genetics (5th ed.). Garland Science.

9. Nurk, S., Koren, S., Rhie, A. et al. (2022). "The Complete Sequence of a Human Genome." Science, 376(6588), 44–53.

10. National Academies of Sciences, Engineering, and Medicine. (2017). Human Genome Editing: Science, Ethics, and Governance. National Academies Press.