Genome-Wides Association Studies (GWAS) Veryan Codd

Why? Identify genetic variants that are associated with a common heritable traits / diseases Understand disease mechanism – identify new genes/pathways involved Personalised medicine – Risk analysis – Early intervention – Inform medication choice

Which genetic changes are studied? SNPs – single nucleotide polymorphisms CNVs – copy number variations Indels - insertions / deletions

Two alleles of a SNP – A and a – Individuals AA, Aa or aa – One allele often less commonly found in population – minor allele (AF<0.5) – Eg rs (RS = reference SNP) MAF=0.27 CC = 0.53, CT=0.40, TT=0.07 SNPs

Linkage Tendency of genetic markers to be inherited together during meiosis. A B C

Linkage – r 2 and D’ D’ is a measure of how frequently two markers are co- inherited R 2 additionally takes into account allele frequencies, is a measure of how highly correlated 2 SNPs are Eg 1 – 2 SNPs, A and B, both have MAF ~50% A B A B A B A B A B A B A B A B A B A B D’=1.0, r 2 =1.0 A would be a “proxy” SNP for B and vice verse – one genotype could fully predict the other

Eg2 2 SNPs, A and B, A has high MAF, B has low MAF A A B A A A A A A A A A A A A A A A A A A D’=1.0 as B can completely predict A r 2 is low - A cannot completely predict B

Genotype individuals on an array – UK Biobank Axiom Array - 820,967 SNP and indel – Imputation to predict additional markers based on known linkage relationships Can test ~10million genetic variants New release of 1000Genomes – 78million SNPs Methods

Analysis…..basics Test the hypothesis that allele frequencies differ between groups e.g. cases vs controls Multiple testing – millions of tests being performed Pragmatic threshold of P=5x10 -8 More recently studies looking at FDR (false discovery rate)

“missing heritability” Even with large scale analysis only ~11% of heritability of CAD explained More small effect common variant to be discovered? Structural variations? Rare variants with large effects? – Exome arrays/ sequencing have been disappointing – Family based studies

Mendelian randomisation Association is NOT causation Genotype assigned randomly from parents to offspring at meiosis Genetics not affected by lifestyle etc which could also impact disease risk Can be affected by – Linkage disequilibrium – Genetic heterogeneity – Pleiotropy – Population stratification

Reasons for an association of a biomarker with a disease Association of TL with CAD TL is involved in disease pathogenesis TL shortened by CAD process Another factor affects both TL and CAD Eg Increased inflammation increases risk of CAD and independently increases TL attrition leading to shorter telomeres CAUSATION REVERSE CAUSATION CONFOUNDING

Mendelian randomisation as a means of establishing causality Schunkert H & Samani NJ. NEJM 2008 Confounding factor (Lifestyle) TL CAD Genetic variant Reverse causation Causal