Overview � Genetic architecture of a trait � Arrays � GWAS � Ethnicity Statistics and analytical issues: � Effect size / power SNP and GWAS � Consortia Linda Broer (l.broer@erasmusmc.nl) Genetic Laboratory Department of Internal Medicine Erasmus MC, Rotterdam Overview Types of genetic studies Genetic architecture of traits � Genetic architecture of a trait rare, monogenic � Arrays (linkage) � GWAS Few examples big � Ethnicity Effect Size � Effect size / power � Consortia common, complex Probably real small (association) (impossible to identify with current methods) rare common Frequency Genetic Variant Modified from McCarthy et al., Nat Genet Rev 2008 1
Genome-wide linkage study Example1: hemophilia in European royalty � Assumption: trait is determined by rare variants with large effect � Hypothesis free � Resolution is poor (5 - 20 million base pairs) � Works well for monogenetic traits � Need to know/estimate model of inheritance! � Common traits / complex diseases? � Not effective Types of genetic studies Candidate gene approach Genetic architecture of traits � Assumption: trait is determined by common variants with small effect rare, monogenic (linkage) � Hypothesis driven Few examples big � Based on prior (biological) knowledge Effect Size � Association analysis of few variants common, complex Probably real small (association) � Excellent resolution (1 bp) (impossible to identify with current methods) � Often results in false-positive or negative findings rare common � Why? Frequency Genetic Variant Modified from McCarthy et al., Nat Genet Rev 2008 2
Example of false-positive candidate gene study Genome-wide approach � Heat Shock Proteins are the most important pathway to determine � Scale-up of candidate gene to genome-wide longevity after IGF1 in model organisms � In centenarians the association between HSP proteins and longevity � Hypothesis free approach shown � In genetics … � Resolution 5-50 thousand base pairs � Very effective Overview Which genotyping technique to use? � Genetic architecture of a trait � Arrays � GWAS � Ethnicity � Effect size / power � Consortia 3
Array-technology for genotyping SNPs Array-technology � Created for genotyping many SNPs (> 0.3 million) � Two major companies: Illumina & Affymetrix � Illumina: tagSNP optimized � Affymetrix: population-specific arrays � Primarily used for Genome-wide testing � GWAS bead Address Probe 23 bp � But also for: pharmacogenetics, clinical research, linkage analysis � Illumina bead-array 50 bp � Beads have probes of one SNP attached � Each bead is spotted in multifold to increase accuracy and redundancy Procedure Procedure � DNA normalization and whole genome amplification � Hybridization on array, single base extension � SBE: 1 base added to the probe SNP Fragmented gDNA Whole genome amplification 200 ng DNA bead Address Probe DNA pellet after bead T-DNP amplification Address Probe Labelled ddNTP 4
Procedure Procedure DNA collection on array Every dot represents a SNP Colors: Red & green: homozygous Yellow: heterozygous Overview GWAS analysis Analyzing all SNPs in 1 run � Genetic architecture of a trait � Arrays Visualizing results in plots � GWAS � Ethnicity � Effect size / power � Consortia Select SNPs Combine GWASs Replication Manhattan-plot Each dot represents 1 SNP Meta-Analysis of all data 5
Manhattan plot: “Holland” plot Manhattan plot: “Dubai” plot LUMBAR SPINE BMD P < 1.10 -206 HERC2/OCA2 gene 12 kb on Chr. 15q11 5 x 10 -8 Rivadeneira et al., Nat Genet., 2009 Rotterdam Study: Kayser et al, Am J Hum Genet, 2008 Manhattan plot: true “Manhattan” plot GWAS catalog (https://www.ebi.ac.uk/gwas/) - 180 loci identified � Online collection of all published GWAS - 10-15% variance explained � Quality controlled � Manually curated � Literature-derived � Regularly updated � Currently contains: � 3,172 publications � 52,491 unique SNP-trait associations 5 x 10 -8 Lango, Estrada, Rivadeneira et al., Nature, 2010 6
GWAS on cardiovascular traits GWAS on cancer 7
Overview Out-of-Africa � Genetic architecture of a trait � Arrays � GWAS � Ethnicity � Effect size / power � Consortia Not all variants got to travel: bottleneck event Not all variants got to travel: bottleneck event � Africans have more variants than Europeans/Asians � ‘Unique’ variants appeared in those that left Africa � Adaptation to new environment � Some of these came from already existing hominids outside Africa � Frequencies of variants can differ between Ethnic groups 8
Side note: humans are not the only species with Example: rs776746 a bottle-neck event � SNP in gene CYP3A5 which metabolizes clinical drugs � Cheetahs � G allele encodes CYP3A5*3 allele � 2 bottle-neck events � Inactivates the gene � 10,000 years ago � Last 100 years � All cheetahs are identical twins � Elephant seals � Only 20-50 individuals left in 1890 � Florida Panthers � Isolated from other cougars � Only 30-50 individuals left in 1980 � Many recessive disease present in population Consequences for study design Overview � Example: � Genetic architecture of a trait � Cases: sickle cell anemia � Arrays � Controls: European ancestry � GWAS � What will you find? � Ethnicity � Multiple variants across the genome show evidence of association � Effect size / power � Most cases are African ancestry � Consortia � All controls are European ancestry 9
Power is an issue in GWAS Effect size and frequency are important to consider TRUTH OR=2 GWA Study H 0 : No H A : Association Association OR=1.5 Accept H 0 Beta ( β) OK No Association error Reject H 0 Alpha ( α ) OK Association error Power (1- β) of a GWA study will depend on: OR=1.3 OR=1.2 FIXED FACTORS MODIFIABLE FACTORS -Allele frequency -Phenotype definition -Effect size -Alpha level 1000 cases / 1000 controls -Linkage disequilibrium -Sample size Power is an issue in GWAS Sample size TRUTH � Sample size needed to detect associations is >>20,000 GWA Study H 0 : No H A : Association � Preferably even over 100,000 samples Association Accept H 0 Beta ( β) OK � Most study populations don’t have this many samples No Association error � Rotterdam Study: ~15,000 samples Reject H 0 Alpha ( α ) OK Association error � Exceptions Power (1- β) of a GWA study will depend on: � UK Biobank: ~500,000 samples FIXED FACTORS MODIFIABLE FACTORS � 23andMe: ~200,000 samples and growing -Allele frequency -Phenotype definition -Effect size -Alpha level � Working together with others is only solution -Linkage disequilibrium -Sample size 10
Overview Large consortia CHARGE � Genetic architecture of a trait � Arrays � GWAS � Ethnicity � Effect size / power Rotterdam � Consortia Study GEnetic Factors of OSteoporosis GENETIC INVESTIGATIONS OF ANTHROPOMETRIC TRAITS Consortia: working together does work � Much larger sample sizes can be achieved � Go from competition to cooperation � Creates better science! � But… � Only ‘cosmopolitan’ variants found � Trying to set up a call with the US, Europe and Australia is impossible � Can slow things down as you are waiting for each other � Typical GWAS takes ~3-7 years 11
LUMBAR SPINE BMD LUMBAR SPINE BMD LRP5 5 x 10 -8 5 x 10 -8 • Rotterdam Study • Rotterdam Study • ERF Study • ERF Study N=5,000 N=6,200 • Twins UK • Twins UK • deCODE Genetics • deCODE Genetics • Framingham Study Rivadeneira et al., Nat Genet., 2009 • Framingham Study Rivadeneira et al., Nat Genet., 2009 LUMBAR SPINE BMD LUMBAR SPINE BMD RANK-L C6ôrf10 OPG 1p36 LRP5 LRP5 5 x 10 -8 5 x 10 -8 MHC • Rotterdam Study • Rotterdam Study • ERF Study • ERF Study N=8,500 N=15,000 • Twins UK • Twins UK • deCODE Genetics • deCODE Genetics • Framingham Study Rivadeneira et al., Nat Genet., 2009 • Framingham Study Rivadeneira et al., Nat Genet., 2009 12
The success of consortia (2005): Everyone doing their own thing LUMBAR SPINE BMD RANK-L C6ôrf10 OPG 1p36 SP7 LRP5 5 x 10 -8 • Rotterdam Study • ERF Study N=19,125 • Twins UK • deCODE Genetics • Framingham Study Rivadeneira et al., Nat Genet., 2009 The success of consortia (2007): starting to work The success of consortia (2009): we’re getting together somewhere 13
The success of consortia (2013) : I’ve given up to count The success of consortia (2011): is anything not them significant? The success of consortia (2015): Wow, it’s pretty ☺ ☺ ☺ ☺ What has/will GWAS achieve E D. Green et al. Nature 470 , 204-213 (2011) doi:10.1038/nature09764 14
In summary / Take Home Messages Questions � Before doing genetic research, determine the genetic architecture of your trait and adjust methodology accordingly � Arrays quickly becoming so cheap that they are feasible for any study � GWAS is the work-horse of genetic epidemiology of complex traits � Allele frequencies (and trait variation) can differ between ethnicities � Sample size is only truly adjustable determinant of power � Working together in consortia not just a necessity, it pays off 15
Recommend
More recommend
