leveraging distant relatedness to quantify human mutation ... · leveraging distant relatedness to...

Leveraging distant relatedness to quantify human mutation and gene

conversion rates

Pier Palamara

Harvard T. H. Chan School of Public Health

ASHG 2015.10.10

Methods for inferring the mutation rate

see e.g. [Scally & Durbin Nat Rev Gen 2012, Segurel et al. Annu. Rev. Genom. Hum. Gen. 2014 ]

generations

[Elango et al. PNAS 2006, Scally et al. Nature 2012]

[image: Pääbo, Nature 2003]


generations

100,000s Phylogenetic methods

generations

100,000s

1

Phylogenetic methods

Trios

[Conrad et al. Nat. Gen. 2011, Sun et al. Nat. Gen. 2012,

Kong et al. Nature 2012, Neale et al. Nature 2012]




[Conrad et al. Nat. Gen. 2011, Sun et al. Nat. Gen. 2012,

Kong et al. Nature 2012, Neale et al. Nature 2012]



Different estimates: 2.4 x 10-8 vs 1.2 x 10-8


generations

100,000s

1


Trios

[Scally & Durbin, Nat. Rev. Gen. 2012, Segurel et al. Annu. Rev. Genom. Hum. Gen. 2014 ]

1,000s Deep genealogical relationships

e.g. [Lipson et al. PLOS Gen. 2015 (in press)] [Image: Tishkoff and Verrelli, 2003]


generations

100,000s

1


Trios

10s Recent genealogical relationships


generations

100,000s

1


Trios

1,000s Deep genealogical relationships

this work

Identity By Descent

IBD

see e.g. [Browning & Browning, Annual Review of Genetics 2012]

Inferring mutation rate in “unrelated” individuals IB

D m

ism

atch

ing

rate

2 × IBD segment age

• tMRCA regression: Regress IBD sequence mismatching rate on age of segments.

Inferring mutation rate in “unrelated” individuals


IBD

mis

mat

chin

g ra

te



slope = mutation rate


2 × TMRCA

IBD

mis

mat

chin

g ra

te


Genotyping errors

slope = mutation rate

Intercept ≈ genotype error


2 × TMRCA

IBD

mis

mat

chin

g ra

te



Inferring the age of IBD segments

? ?

Unknown TMRCA Infer from demographic history

[Palamara et al. AJHG 2012] [Ralph & Coop, PLOS Bio. 2013]

IBD

• Gene conversion occurs at a rate proportional to recombination

• When it occurs, an existing SNP may be copied on IBD haplotypes

Dealing with non-crossover gene conversion

Gene conversion

May harbor common variant


• When it occurs, an existing SNP may be copied on IBD haplotype


… with probability proportional to number of generations

2 × TMRCA

IBD

mis

mat

chin

g ra

te



• When it occurs, an existing SNP may be copied on IBD haplotype


… with probability proportional to number of generations and variant frequency…

2 × TMRCA

IBD

mis

mat

chin

g ra

te


• Solution: perform a second regression, now using threshold on maximum MAF variants in sequence

Any polymorphic variant used

Maximum MAF

Infe

rred

mu

tati

on

rat

e

Non-crossover gene conversion: MaAF regression


Only variants with MAF < 0.4

Maximum MAF

Infe

rred

mu

tati

on

rat

e

Maximum MAF



Gene conversion-corrected estimate

Maximum MAF

Infe

rred

mu

tati

on

rat

e

Maximum MAF



proportional to gene conversion

If population heterozygosity is known, can infer rate of gene conversion

Maximum MAF

Infe

rred

mu

tati

on

rat

e

Maximum MAF


• Results: simulation and real data

tMRCA regression is robust to genotyping error

Simulated error rate

Infe

rred

mu

tati

on

rat

e (x

10

8)

IBD approach is more efficient than trio approach

Trios

IBD in GoNL

25 50 100 200 400

5e−10

1e−09

2e−09

5e−09

Samples

Sta

nda

rd e

rror

of estim

ate

Sample size

Stan

dar

d e

rro

r o

f es

tim

ate

Real data: the Genome of the Netherlands

• ~250 trios1

• ~13x coverage (~26x on transmitted haplotype)

• Trio-phased using MVNcall2

• IBD detected using GERMLINE3 (+ filtering)

• Demographic history (piece-wise expansion)

inferred using DoRIS4

Two periods of exponential expansion

Stronger recent expansion (Golden age)

1: [Francioli et al., Nat. Gen. 2014] 2: [Melanou & Marchini, Bioinformatics 2013] 3: [Gusev et al., Gen. Res. 2009] 4: [Palamara et al., AJHG 2012]

• When gene conversion correction is applied, for segments > 1.6cM, μ = 1.66 x 10-8, s.e. 0.04 x 10-8

• Higher than pedigree-based μ

Inferring μ in real data

Gene conversion-corrected estimate

Maximum MAF

Infe

rred

mu

tati

on

rat

e


• Gene conversion rate of 5.99 x 10-6, s.e. 0.69 x 10-6

(Matches estimate of Williams et al. eLife 2015)

Inferring gene conversion rate in real data


• Gene conversion rate of 5.99 x 10-6, s.e. 0.69 x 10-6

(Matches estimate of Williams et al. eLife 2015)

• Same method can be applied to estimate rate of short indels

μindel = 1.26 x 10-9, s.e. 0.06 x 10-9

(Compatible with Besenbacher et al. Nat. Comm. 2015)

Inferring indel rate in real data

• Rec. and mut. rates strongly correlated (p<10-5)

• After controlling for gene conversion, no association (p=0.17)

Recombination Mutation

Local standardized recombination rate

Infe

rred

mu

tati

on

rat

e

B statistic closely reflects local IBD sharing (p<10-6) But no impact on mutation rate estimate (p=0.19)

Selection Mutation

Local B statistic

B statistic: [McVicker et al. PLOS Gen. 2009]

Ave

rage

IBD

seg

men

t le

ngt

h

• Mismatching variants on IBD enriched for deleterious variation

• No evidence for enrichment/depletion of mutation rate in several genomic annotations

Other analyses

Conclusions and future work

• New method to infer mutation and gene conversion rates

– μ = 1.66 x 10-8 (higher than pedigree studies) • Agrees with recent estimate of Lipson et al. PLOS Gen. 2015 (in press)

– No effects of recombination/selection on estimate

– No enrichment/depletion in functional annotations

• Use in multi-generation pedigree data

Acknowledgements

Funding

NIH R01 MH101244, U54 CA121852-06; NSF 08929882, 0845677

PauI de Bakker Cisca Wijmenga

Clara C Elbers Sara L Pulit Androniki Menelaou Laurent Francioli

Genome of the Netherlands consortium

Abdel Abdellaoui Albert Hofman Alexandros Kanterakis Andre G Uitterlinden Anton JM de Craen Ben Oostra Bruce H Wolffenbuttel Cornelia M van Duijn Dorret Boomsma Eka HD Suchiman Eline P Slagboom Fernanodo Rivadeneira

Freerk van Dijk Gert-Jan van Ommen Gonneke Willemsen Heorhiy Byelas Hongzhi Cao Jeanine Houwing-Duistermaat Itsik Pe'er Jeroen FJ Laros Jessica van Setten Johan den Dunnen Jouke Jan Hottenga Jun Wang

Kai Ye Karol Estrada Lennart C Karssen Marian Beekman Martijn Dijkstra Martijn Vermaat Mathijs Kattenberg Morris A Swertz Ning Li Paz Polak Peter de Knijff Pier Palamara

Pieter B Neerincx Qibin Li Ruoyan Chen Shamil Sunyaev Sujie Cao Victor Guryev Vyacheslav Koval Wigard Kloosterman Yingrui Li Yuanping Du

Palamara et al. AJHG 2015 (in press) available on BioRxiv

(Harvard University) Alkes Price, John Wakeley, Shamil Sunyaev, Alexander Gusev, Peter Wilton, Hilary Finucane, Sriram Sankararaman

(University Medical Center Utrecht ) Laurent Francioli, Paul de Bakker

(Columbia University) Itsik Pe’er

(Broad Institute) Giulio Genovese

Genotyping error is captured by intercept

IBD approach is more efficient than trio approach

● ● ●Trios GoNL MASAI

50 100 200

2e−10

5e−10

1e−09

2e−09

5e−09

1e−08

2e−08

Samples

Sta

nda

rd e

rror

of estim

ate

No effects of background selection on inference

●

●

● ●

●

●

●●

●

●

●●

●

−5 −4 −3 −2 −1 0 1

1.0

e−

08

1.5

e−

08

2.0

e−

08

2.5

e−

08

Standardized B statistic in region

Infe

rre

d m

uta

tion

rate

0 100 200 300 400

Time (generations)

Effective

siz

e

10

310

41

05

10

6● ● ● ●Ashkenazi European Masai Dutch

0.0 0.2 0.4 0.6 0.8 1.0

0.001

0.002

0.005

0.010

0.020

0.050

0.100

0.200

Derived allele frequency

Pro

bab

ility

● ● ●alpha = 0.01 alpha = 0.5 alpha = 1

5 10 15 20

Physical position (Mb)

cM

/Mb

0

2

4

6

8 ● ●steps map hotspots map

● ● ● ● ● ●●

●●

●●

●

●

●

●

●

●

●

●

●

●

Posterior threshold

Inte

rce

pt

0.0 0.2 0.4 0.6 0.8 1.0

2.0e−06

2.5e−06

3.0e−06

3.5e−06

4.0e−06

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

Posterior threshold

Infe

rred

muta

tio

n r

ate

0.0 0.2 0.4 0.6 0.8 1.0

1.4e−08

1.5e−08

1.6e−08

1.7e−08

1.8e−08

1.9e−08

●

●

●

●

●●

● ● ● ● ● ● ● ● ● ● ● ● ●● ●

●

●

●

●

●

●●

●●

●●

● ● ● ● ● ● ●●

●

Minimum IBD segment length (cM)

Infe

rred

muta

tio

n r

ate

1.0 1.5 2.0 2.5

1.5e−08

2.0e−08

2.5e−08

3.0e−08

●

●

No correction

With gene conversion correction

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

q = −2.719e−05 + 1.480e−03 F

r2

= 0.9964

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

q = −2.436e−06 + 1.170e−04 F

r2

= 0.9966

Maximum allowed frequency ( F )

Hete

rozygo

sity (

q )

0.0 0.1 0.2 0.3 0.4 0.5

0e+00

2e−04

4e−04

6e−04

8e−04 ●

●

Point mutations

Indels

●●

●

● ●

●

●

●

●

●

●

0.0e+00 5.0e−08 1.0e−07 1.5e−07 2.0e−07 2.5e−07 3.0e−07

0.0

00

00

0.0

00

05

0.0

00

10

0.0

00

15

0.0

00

20

Average recombination rate in region

Infe

rre

d g

en

e c

onve

rsio

n r

ate

pe

r b

p

leveraging distant relatedness to quantify human mutation ... · leveraging distant relatedness to...

Documents