Inferring human population history from multiple genome sequences

Size: px

Start display at page:

Download "Inferring human population history from multiple genome sequences"

Pamela Park
5 years ago
Views:

1 Inferring human population history from multiple genome sequences Stephan Schiffels Postdoctoral Fellow with Richard Durbin, Wellcome Trust Sanger Institute, Cambridge, UK

2 From genome sequences to human history CEU MXL TSI GIH CHB JPT YRI MKK LWK [Sequence data from Complete Genomics]

From genome sequences to human history time [kya] 200 100 50 20 10 CEU

3 From genome sequences to human history time [kya] CEU TSI GIH JPT CHB MXL MKK LWK YRI [Sequence data from Complete Genomics]

4 How are sequences related? Ancestral recombinations change trees along the sequences time (past) Haplotypes ACCTG... ACCTG... ACCTG... ACCTG... Mutations Problem: Estimate trees only from observed mutations

5 Previous work: PSMC [Li and Durbin, 2011]

6 Previous work: PSMC years ago [Li and Durbin, 2011]

7 Previous work: PSMC years ago But: Inference from only two sequences limited to times beyond 20kya. Also: population splits difficult to model [Li and Durbin, 2011]

8 Multiple sequentially Markovian Coalescent (MSMC) time (past) First Coalescence t (hidden state) Haplotypes ACCTG... ACCTG... ACCTG... ACCTG... Mutations time MSMC hidden state: triple (t, i, j) t i j

9 Effect of recombination on Genealogies time x x Ancestral recombination Cut branch Re-coalesce New tree [Sequentially Markovian Coalescent, McVean and Cardin, 2005]

10 MSMC: state transitions (s,k,l) (t,i,j) where t<s s t<s k l i j

11 MSMC: state transitions (s,k,l) (t,i,j) where t<s s t<s k l i j (s,k,l) (t,i,j) where t=s s t=s or t=s k l i j i j

12 MSMC: state transitions (s,k,l) (t,i,j) where t<s s t<s k l i j (s,k,l) (t,i,j) where t=s s t=s or t=s k l i j i j (s,k,l) (t,i,j) s t>s where t>s k l i j

13 MSMC: mutation probability Total Branchlength T Singleton Branchlength TS First Coalescence t No mutation 1 - µt A A A A Singleton within pair µ t A C A A Singleton outside pair µ A A A T Double mutation 0 A T A T Higher Frequency µ A T T A Missing Data 1 A A - -

14 Local Inference of first coalescence time 2 haplotypes, similar to PSMC [Li and Durbin, 2011] 4 haplotypes 8 haplotypes

15 Test with simulations Example: Exponentially growing and shrinking population size effective population size Simulation 2 hapl. 4 hapl. 8 hapl time [years ago]

16 Test with simulations Example: Exponentially growing and shrinking population size effective population size Simulation 2 hapl. 4 hapl. 8 hapl haplotypes: 40kya-3Mya time [years ago]

17 Test with simulations Example: Exponentially growing and shrinking population size effective population size Simulation 2 hapl. 4 hapl. 8 hapl time [years ago] 2 haplotypes: 40kya-3Mya 4 haplotypes: 8kya-300kya

18 Test with simulations Example: Exponentially growing and shrinking population size effective population size Simulation 2 hapl. 4 hapl. 8 hapl haplotypes: 40kya-3Mya time [years ago] 4 haplotypes: 8kya-300kya 8 haplotypes: 2kya-80kya

19 From genome sequences to human history CEU MXL TSI GIH CHB JPT YRI MKK LWK [Sequence data from Complete Genomics]

20 Inferring historical Population Sizes real time scaling using mutation rate per generation µ= and a generation time of 30 years

21 Inferring historical Population Sizes 10 6 YRI (Nigeria) MKK (Kenya) LWK (Kenya) CEU (N. Europe) TSI (Italy) CHB (China) JPT (Japan) GIH (N. India) MXL (only Nat. Am.) PEL (only Nat. Am.) real time scaling using mutation rate per generation µ= and a generation time of 30 years Effective population size Mild bottleneck in African ancestors time [years]

22 Inferring historical Population Sizes 10 6 YRI (Nigeria) MKK (Kenya) LWK (Kenya) CEU (N. Europe) TSI (Italy) CHB (China) JPT (Japan) GIH (N. India) MXL (only Nat. Am.) PEL (only Nat. Am.) real time scaling using mutation rate per generation µ= and a generation time of 30 years Effective population size Mild bottleneck in African ancestors 10 3 identical history of non- African populations before 50kya time [years]

23 Inferring historical Population Sizes 10 6 YRI (Nigeria) MKK (Kenya) LWK (Kenya) CEU (N. Europe) TSI (Italy) CHB (China) JPT (Japan) GIH (N. India) MXL (only Nat. Am.) PEL (only Nat. Am.) real time scaling using mutation rate per generation µ= and a generation time of 30 years Effective population size Mild bottleneck in African ancestors 10 3 identical history of non- African populations before 50kya time [years] strong bottleneck in non- African populations 50-60kya, coincides with out-of-africa dispersal

24 Inferring historical Population Sizes 10 6 YRI (Nigeria) MKK (Kenya) LWK (Kenya) CEU (N. Europe) TSI (Italy) CHB (China) JPT (Japan) GIH (N. India) MXL (only Nat. Am.) PEL (only Nat. Am.) real time scaling using mutation rate per generation µ= and a generation time of 30 years Effective population size Mild bottleneck in African ancestors 10 3 identical history of non- African populations before 50kya time [years] deep bottleneck in Native American populations around 15kya: Beginning of the Peopling of America via Alaskan Landbridge strong bottleneck in non- African populations 50-60kya, coincides with out-of-africa dispersal

25 Inferring historical Population Sizes 10 6 YRI (Nigeria) MKK (Kenya) LWK (Kenya) CEU (N. Europe) TSI (Italy) CHB (China) JPT (Japan) GIH (N. India) MXL (only Nat. Am.) PEL (only Nat. Am.) real time scaling using mutation rate per generation µ= and a generation time of 30 years Effective population size Mild bottleneck in African ancestors 10 3 identical history of non- African populations before 50kya Complex pattern of neolithic expansions in the last 10,000 years within Europe, Asia and Africa time [years] deep bottleneck in Native American populations around 15kya: Beginning of the Peopling of America via Alaskan Landbridge strong bottleneck in non- African populations 50-60kya, coincides with out-of-africa dispersal

26 Population size inference from 8 haplotypes

27 Population size inference from 8 haplotypes 10 8 CEU (N. Europe) TSI (Italy) CHB (China) JPT (Japan) GIH (N. India) YRI (Nigeria) MKK (Kenya) LWK (Kenya) 10 7 effective population size time [years ago] Bantu Expansion within Africa?

28 Population size inference from 8 haplotypes 10 8 CEU (N. Europe) TSI (Italy) CHB (China) JPT (Japan) GIH (N. India) YRI (Nigeria) MKK (Kenya) LWK (Kenya) 10 7 effective population size Bantu Expansion within Africa? time [years ago] CEU: Constant population size until 2kya, then expansion Structure and admixture in Southern Europe [see Botigué et al., 2013]

29 Population size inference from 8 haplotypes 10 8 CEU (N. Europe) TSI (Italy) CHB (China) JPT (Japan) GIH (N. India) YRI (Nigeria) MKK (Kenya) LWK (Kenya) effective population size Continuous expansion in Chinese ancestors Bantu Expansion within Africa? time [years ago] CEU: Constant population size until 2kya, then expansion Structure and admixture in Southern Europe [see Botigué et al., 2013]

30 Divergence between populations Idea: Infer separate coalescence rates within and between populations: First Coalescence within Population 1 First Coalescence within Population 2 First Coalescence across both populations MSMC can infer separate coalescence rates within populations, Given rates within populations, λ11(t) and λ22(t), and across populations, λ12(t), compute relative gene flow as ratio m(t) = λ12(t) [λ11(t)+ λ22(t)] / 2

31 Testing gene flow inference with simulated split gene flow kya split sim. 10kya split, 4 hapl. 10kya split, 8 hapl. 100kya split sim. 100kya split, 4 hapl. 100kya split, 8 hapl. m(t)=1: perfectly mixed time [years ago] m(t)=0: perfectly split 4 haplotypes: good for splits kya. 8 haplotypes: good for splits 5-50kya.

32 African Population separations

33 African Population separations relative gene flow MXL/YRI CEU/YRI CHB/YRI CEU/MKK CEU/LWK YRI/MKK LWK/MKK YRI/LWK African(YRI)/Non- African separation: 150kya-40kya time [years ago] 8 haplotypes

34 African Population separations relative gene flow MXL/YRI CEU/YRI CHB/YRI CEU/MKK CEU/LWK YRI/MKK LWK/MKK YRI/LWK African(YRI)/Non- African separation: 150kya-40kya time [years ago] 8 haplotypes 1.0 relative gene flow time [years] CEU/YRI 150kya split sim. 100kya split sim. 50kya split sim

35 African Population separations relative gene flow MXL/YRI CEU/YRI CHB/YRI CEU/MKK CEU/LWK YRI/MKK LWK/MKK YRI/LWK African(YRI)/Non- African separation: 150kya-40kya time [years ago] 8 haplotypes Masaai remained mixing with European ancestors until ~80kya relative gene flow CEU/YRI 150kya split sim. 100kya split sim. 50kya split sim time [years]

36 African Population separations relative gene flow MXL/YRI CEU/YRI CHB/YRI CEU/MKK CEU/LWK YRI/MKK LWK/MKK YRI/LWK African(YRI)/Non- African separation: 150kya-40kya time [years ago] 8 haplotypes Complex splits within Africa: 5kya-30kya Masaai remained mixing with European ancestors until ~80kya relative gene flow CEU/YRI 150kya split sim. 100kya split sim. 50kya split sim time [years]

37 Non-African Population Separations

38 Non-African Population Separations relative gene flow CHB/CEU MXL/CEU CHB/MXL GIH/MXL CHB/GIH GIH/CEU CHB/JPT CEU/TSI 8 haplotypes time [years ago] European/East-Asian split: 30kya, small older component?

39 Non-African Population Separations relative gene flow CHB/CEU MXL/CEU CHB/MXL GIH/MXL CHB/GIH GIH/CEU CHB/JPT CEU/TSI 8 haplotypes time [years ago] Asian/American split: 20kya European/East-Asian split: 30kya, small older component?

40 Non-African Population Separations relative gene flow CHB/CEU MXL/CEU CHB/MXL GIH/MXL CHB/GIH GIH/CEU CHB/JPT CEU/TSI 8 haplotypes time [years ago] Complex Indian separations between 8kya and 20kya Asian/American split: 20kya European/East-Asian split: 30kya, small older component?

41 Non-African Population Separations relative gene flow CHB/CEU MXL/CEU CHB/MXL GIH/MXL CHB/GIH GIH/CEU CHB/JPT CEU/TSI 8 haplotypes Splits within Europe and within East-Asia Complex Indian separations between 8kya and 20kya time [years ago] Asian/American split: 20kya European/East-Asian split: 30kya, small older component?

42 Implications of higher human mutation rate mutation rate per generation µ= relative gene flow Nat.Am./Asn. split: 10kya??? Afr/Non-Afr. split: 35kya??? 0.2 MXL/YRI CEU/YRI CHB/YRI CHB/MXL GIH/MXL time [years ago] A higher mutation rate of would push these splits towards too recent times

43 MSMC Summary on separation history time [kya] CEU TSI GIH JPT CHB MXL MKK LWK YRI [Schiffels and Durbin, Nature Genetics, in press]

44 Acknowledgements Richard Durbin Aylwyn Scally Jeff Kidd, Simon Gravel, Eimear Kenny, Carlos Bustamante for MXL and PEL admixture tracts Article in press: Schiffels and Durbin, Nature Genetics Preprint available on biorxiv. Software available on

45 Transition probability Probability to remain in state (i,j,t) Probability to change time (and pair) of first coalescence depends on: coalescence rates λ(t) recombination rate r Number of sequences M

Human Ancestry (Learning Objectives)

Human Ancestry (Learning Objectives) 1. Identify the characters shared by all primates and relate them to the function they served in their common ancestor. 2. Learn the fields study of Human evolution