Pop Gen meets Quant Gen and other open questions

Transcription

1 Pop Gen meets Quant Gen and other open questions Ryan D. Hernandez Tim O Connor ryan.hernandez@ucsf.edu 1

2 Modern Human Genomics 2

3 Human Colonization of the World Kennewick 9,500 years ago Spirit Cave 9,500-9,400 years ago 6 20,000-15,000 years ago NORTH AMERICA Yana River 30,000 years ago Clovis 13,500 years ago Meadowcroft 19,000-12,000 years ago 40,000 years ago 5 Zhoukoudian (Shandingdong) 11,000 years ago 40,000-30,000 years ago 4 EUROPE Pestera cu Oase 35,000 years ago ASIA Minatogawa 18,000 years ago AFRICA Nile River 1 Qafzeh 100,000 years ago Red Sea 2 Omo Kibish Oldest modern human 195,000 years ago 200,000 years ago 70,000-50,000 years ago Andaman Islands EQUATOR Niah Cave 40,000 years ago SOUTH AMERICA Malakunanja 50,000 years ago 15,000-12,000 years ago Monte Verde 14,800 years ago Human Migration Fossil or artifact site Klasies River Mouth 120,000 years ago 40,000 years ago Migration date Generalized route SOURCES: SUSAN ANTON, NEW YORK UNIVERSITY; ALISON BROOKS, GEORGE WASHINGTON UNIVERSITY; PETER FORSTER, UNIVERSITY OF CAMBRIDGE; JAMES F. O'CONNELL, UNIVERSITY OF UTAH; STEPHEN OPPENHEIMER, OXFORD UNIVERSITY; SPENCER WELLS, NATIONAL GEOGRAPHIC SOCIETY; OFER BAR-YOSEF, HARVARD UNIVERSITY NGM MAPS AUSTRALIA 3 Lake Mungo 45,000 years ago 50,000 years ago 2006 National Geographic Society. All rights reserved. 3

4 Heritability and Human Height Studies of heritability ask questions such as how much genetic factors play a role in differences in height between people. This is not the same as asking how much genetic factors influence height in any one person

5 An estimated 80% of variation in height driven is driven by genetics Large twin study Silventoinen et al, 2003 Twin Research 5

6 But GWAS explain only 20% of the variation in height The narrow-sense heritability 2 by summing the hgwas : explained effects of GWAS identified SNPs. 250,000 subjects 6 Wood et al, 2014 Nat. Genet. i.ytimg.com/vi/e0aeks_id6c/maxresdefault.jpg

7 GWAS have the potential to explain 60% of the variation in height 2 The narrow-sense heritability hg : explained by all genotyped SNPs. 250,000 subjects 7 Wood et al, 2014 Nat. Genet. i.ytimg.com/vi/e0aeks_id6c/maxresdefault.jpg

8 Challenges For Studying Complex Diseases Sit 8 The case of the missing heritability Maher, Nature (2008).

9 Fig. 4. Signals of polygenic adap tsds of SNPs, where tsds > 0 i frequency of the tall allele in based study (20). The x axis is or significant SNPs (P ~ 1)to most s and SNPs are placed into bins o tive SNPs for easier visualization. ( height Z score and SDS, as a func provides evidence that selection polygenic (P = ;LDsc (C)QQ-plottestingforacorrelatio Z score and tsds for 43 traits. t increased frequency of the trait- Significant traits that are also nom by LD score regression (P <0.05 are labeled. SDS replicates signature of selection on height RESEARCH REPORTS 9

10 Major Problem There are no complex traits in which we know: The number of causal variants The frequencies of all the causal variants The effect sizes of all the causal variants The fitness effect of all the causal variants We need a thorough simulation study where we can vary all of these parameters and see how they effect our answer! 10

11 Possible Origins Of Missing Heritability Candidates Common variants of weak effect Incomplete linkage to causal alleles/multiple causal alleles in locus GxG / GxE Interactions Rare variants Structural variation 11

12 From GWAS To Deep Sequencing Genome-wide association studies (GWAS) seek to identify common variants that contribute to common disease Successfully identified many candidate disease-associated genes Challenges: Generally have low relative risk Explain only a small proportion of the phenotypic variance Provides candidate loci, but causal variant is rarely typed Implication: Predictive power of GWAS is minimal 12

13 Missing heritability - calculating variance accounted for by GWAS Distribution of Very Important Phenotype Suppose k variants are found to be associated with VIP Frequency Contribution from each SNP Total variance from GWAS VIP value (arbitrary units) Compare to GWAS 13 Lawrence Uricchio

14 Where is the missing heritability? 14

15 Population Genetics Why would cases have an excess of rare non-synonymous variants in disease-associated genes? Recent neutral mutations that have not had time to spread Deleterious mutations restricted to low frequency Population genetic analyses are ideally suited to distinguish these cases. 15

16 Evolutionary Models Of Complex Disease SNP Disease propensity Disease Fitness Direct relationship between disease and fitness

17 Evolutionary Models Of Complex Disease SNP Disease propensity Other Phenotype Disease Fitness Pleiotropy: SNP impacts multiple phenotypes Uricchio et al., Genome Research (2016)

18 The Model Of Eyre-Walker (2010) The phenotypic effect size has a direct relationship to selection coefficient of causal mutations: z = S (1 + ) Where: ε ~ N(0, σ 2 ) δ = random sign (trait increasing/decreasing) S = selection coefficient τ = measures how the mean absolute effect of a mutation on the trait increases with the strength of selection Eyre-Walker, PNAS (2010)

19 The Model Of Eyre-Walker (2010) As τ decreases, common alleles play a larger role in the phenotype because the effect sizes of weakly deleterious alleles in- crease relative to strongly deleterious alleles. Eyre-Walker, PNAS (2010)

20 The Model Of Simons Et Al (2014) The phenotypic effect size may have a direct relationship to selection coefficient of causal mutations: Where: ρ = Probability that the trait effect is proportional to the selection coefficient: Pleiotropy!! s = selection coefficient sr = random selection coefficient Simons et al, Nat Genet (2014)

21 The Model Of Uricchio Et Al (2016) A hybrid of the two: Where: δ = random sign (trait increasing/decreasing) τ = measures how the mean absolute effect of a mutation on the trait increases with the strength of selection ρ = Probability that the trait effect is proportional to the selection coefficient: Pleiotropy!! s = selection coefficient sr = random selection coefficient Uricchio et al, Genome Research (2016)

22 Evolutionary Models Of Complex Disease SNP Disease propensity Other Phenotype Disease Fitness ρ: τ: correlation(effect size, fitness) (Simons et al, 2014) transforms fitness effect to phenotype (Eyre-Walker, 2010) Pleiotropy: SNP impacts multiple phenotypes Uricchio et al., Genome Research (2016)

23 Why should we think about evolution? Selection pressure towards an optimum Trait optimum Phenotype distribution 23

24 Stabilizing selection Selection pressure towards an optimum Trait optimum Phenotype distribution 24

25 Stabilizing selection Trait optimum Selection pressure towards an optimum Phenotype distribution 25

26 Stabilizing selection Selection pressure towards an optimum Trait optimum New mutations deleterious Larger effect mutations are more deleterious Effect sizes may not be linear in selection strength Want to allow for pleiotropy Phenotype distribution 26

27 A model for selection & effect size τ : selection coefficient effect size 0.15 ρ : selection coefficient effect size 27

28 Human-specific demography and Selection Growth model: Gutenkunst et al (2009) Explosive growth: Tennessen et al (2012) Fitness effects in non-coding DNA: Torgerson et al (2009) Synonymous Neutral: 10 5 < s < 0 Nearly Neut.: 10 4 < s < 10 5 Weak: 10 3 < s < 10 4 Moderate: 10 2 < s < 10 3 Strong: s < AFRICA EUROPE ASIA , , % % 0.5 1% 1 2% 2 5% 5 10% % effect size = f(demography, natural selection) 28 Uricchio, et al. Genome Res 26, (2016).

29 Neutral model: most variance explained by common alleles Standard Neutral Model Proportion of variance explained by alleles with freq ω Ultra rare Intermediate rare Common 5e-04 1e-03 5e-03 1e-02 5e-02 1e-01 5e-01 1e+00 derived allele frequency, ω 29 Uricchio, et al. Genome Res 26, (2016).

30 Genetic architecture is altered by selection and demography AFR, Growth V ω V e-04 5e-03 5e-02 5e-01 derived allele frequency, ω Uricchio, et al. Genome Res 26, (2016).

31 Genetic architecture is altered by selection and demography AFR, Growth V ω V e-04 5e-03 5e-02 5e-01 derived allele frequency, ω Uricchio, et al. Genome Res 26, (2016).

32 Genetic architecture is altered by selection and demography Uricchio, et al. Genome Res 26, (2016).

33 Genetic architecture is altered by selection and demography Implication: in some cases, largest effect alleles are very rare, so we may not detect them with GWAS! Uricchio, et al. Genome Res 26, (2016).

34 Demography and selection matter! As populations expand and contract, or strength of selection changes, the frequency spectrum responds. This can and should impact the genetic architecture of traits! Proportion of variants that are singletons Proportion of trait variance due to singletons Uricchio, et al. Genome Res 26, (2016).

35 Demography and selection matter! Demography and selection also impacts the number of causal variants! Isolated European African Lohmueller, PLoS Genet (2014).

36 Demography and selection matter! Demography and selection themselves do not impact the heritability of traits! Isolated European African

37 The phenotypic legacy of admixture between modern humans and Neandertals We discussed admixture Non-African individuals have ~1-4% Neanderthal ancestry in their genomes. What is it doing? Analysis: 1000 electronic health record (EHR) derived phenotypes in ~28,000 adults of European ancestry 37 Simonti, et al., Science (2016)

38 Downloaded from on July 18, 2017 The phenotypic legacy of admixture between modern humans and Neandertals Fig. 1. Analysis of EHRs reveals clinical effects of Neandertal alleles 38 trait associations, we performed a discovery meta-analysis across emerge E1 Simonti, et al., Science (2016)

39 The phenotypic legacy of admixture between modern humans and Neandertals Phenotype Replication Discovery (E1) Replication (E2) (E2; two-grm) Risk explained P Risk explained P Risk explained P Actinic keratosis 0.64% % % Mood disorders 1.11% % % Depression 2.03% % % Obesity 0.59% % % Seborrheic keratosis 0.77% % % Overweight 0.60% % % Acute upper respiratory infections 0.70% % % Coronary atherosclerosis 0.68% % % Discovery Replication Phenotype Chr:position (hg19) SNP Flanking gene(s) Odds ratio P Odds ratio P Hypercoagulable 1: rs SELP state... Protein-calorie malnutrition 1: rs SLC35F Symptoms involving urinary system 11: rs RHOG, STIM Tobacco use disorder 3: rs SLC6A Simonti, et al., Science (2016)

40 Open Questions What does does the genetic architecture of a complex trait really look like? How many causal variants are there? Proportion of effects from rare/common alleles? Additive vs epistatic interactions? Pleiotropy? 40