QTL Mapping in Heterogeneous Stocks Talbot et al, Nature Genetics (1999) 21:305-308 Mott et at, PNAS (2000) 97:12649-12654.

QTL Mapping in Heterogeneous Stocks

•Talbot et al , Nature Genetics (1999) 21:305-308•Mott et at, PNAS (2000) 97:12649-12654

0

2

4

6

8

10

12

0 10 20 30 40cM

LO

D S

co

re

0

2

4

6

8

10

12

14

16

0 20 40 60 80 100 120cM

LO

D S

co

re

Chromosome 1 Chromosome 15

QTL Detection in F2 to get 30cM resolution

0

2

4

6

8

10

12

14

16

45 50 55 60 65 70 75

cM

LO

D s

co

reChromosome 10

Gershenfeld et al Behav Genetics, 1997

Heterogeneous Stocks (HS)

• A murine cross formed from 8 inbred founders

• Randomly outbred for >60 generations, • 40 mating pairs per generation• Each chromosome is a random mosaic of

the founders• Average distance between recombinants is

1/60=1.7cM

Fine Mapping with HS

QTL

Mosaic of progenitor strains

Region Scan with microsatellites or SNPs.

Test for association between marker and trait by ANOVA

Mapping by Single-Marker Association

Look for association between the phenotype and each marker in the genome scan:

• A marker m has alleles 1…k• Assume phenotypic effect for allele a is Vma • phenotype for individual with genotype a,b is

Vma + Vmb

• Estimate Vma ‘s by analysis of variance• Conclude QTL linked to marker m if some Vma are

significantly different

Single-marker QTL mapping

• Genotype and phenotype ~750 HS mice over 5 regions where QTL detection indicated the presence of a QTL

• Test for the association between the phenotype and marker allele as with F2 cross

• Two QTL fine-mapped to <1cM

HS provides High Resolution

0

1

2

3

4

5

6

7

8

9

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Distance (cM)

D1Mit264

D1Mit394

D1Imm103

D1Mit100

D1Mit423

D1Mit198D1Mit194

D1Mit102

D1Mit289

D1Mit369

-lo

g P

val

ue

Failures

0

1

2

3

4

5

6

7

8

42 42.5 43 43.5 44 44.5 45 45.50

1

2

3

4

5

6

7

8

67.5 68 68.5 69 69.5 70

Chromosome 10 Chromosome 15

Single-Marker Analysis Can Fail

Only 2/5 QTL detected in the F2 cross were confirmed by SM in the HS mice. Why ?

• Genetic Drift eliminated some of the QTL. Simulations indicate this is unlikely.

Single-Marker Analysis Can Fail

Only 2/5 QTL detected in the F2 cross were confirmed by SM in the HS mice. Why ?

• Chromosomes with the same marker allele may be descended from different strains, and so have different trait effects.

Wrong Phase, No Effect

Marker 1: No effect

observable

Marker 2: Observable

effect

QTL

Single-Marker Analysis Can Fail

Only 2/5 QTL detected in the F2 cross were confirmed by SM in the HS mice. Why ?

• Chromosomes with the same marker allele may be descended from different strains, and so have different trait effects.

• Need to test for association between trait and strain rather than trait and marker

Fine Mapping with HSQTLMosaic of progenitor strains – Hidden Data

Region Scan with microsatellites or SNPs. Observed Genotypes

Loss of information because #alleles < #strains & phase unknown

Must infer progenitors from genotypes and test for presence of QTL in each interval

marker D1MIT498 5 alleles position 64.000

A/J AKR BALB C3H C57 DBA I RIIIallele ND 0.125 0.125 0.125 0.125 0.125 0.125 0.125 0.125allele 132 0.500 0.000 0.000 0.500 0.000 0.000 0.000 0.000allele 155 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000allele 153 0.000 0.000 0.250 0.000 0.250 0.250 0.250 0.000allele 130 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000

m m+1s

t

s’

t’

Pmi(s,t) Qm+1,i(s’,t’)

QTL

dm

cdm

HS Interval Mapping

Comparison of SM and DP QTL localisation (a)

Comparison of SM and DP QTL localisation (b)

Summary(1)

• Heterogeneous Stocks provide an experimental means for fine-mapping QTL of small effect

• Dynamic Programming provides a powerful statistical means for analysing HS data

• 5/5 QTL for behaviour were detected and fine-mapped

• Mott et al (2000) PNAS 97:12649-12654

HAPPY web resources

HAPPY home page:

http://www.well.ox.ac.uk/happy

Web Server:

http://zeno.well.ox.ac.uk:8080/git-bin/happy.cgi

HAPPY is a program to map QTLs in Heterogeneous Stocks

Tea !

Future Directions (1)Mapping Traits in Parallel

• We propose to genotype 3000 markers 1cM apart on 2000 HS mice

• Measure as many phenotypes as possible affecting asthma, diabetes, behaviour etc on these animals

• Map all genes affecting these traits in parallel• Much cheaper than scanning diseases separately• Simulations indicate >90% probability of

detecting any gene accounting for >2.5% of phenotypic variance at genome-wide 5% significance level

Future Directions(2)Mapping Modifier Genes

• Over 2500 mouse models transgenic for human

• On inbred or backcross background

• HS x transgenic hybrid can be used for mapping modifiers

• Need to extend analysis for F2 HS x inbred

Inbred-Outbred Cross

• Detection phase – a genome scan with ~100 markers at 20-30 cM seperation

• Fine-Mapping Phase – rescan at 1cM spacing only those regions which were detected

Simultaneous Detection And Fine Mapping Using an Inbred

Outbred Cross

x

Genome Scan using F2chromosome

Fine Map usingHS chromosome

HS Mouse Inbred strain (knockout) F2

Inbred-Outbred Analysis

• QTL detection depends on variance between HS and background

• Fine-mapping depends on variance within HS

• Power depends on how total variance is split between the detection and fine-map

• phases

Power depends on the modifier allele frequency

Variance under Dominance Model

detection

full

difference

HS

Proportion of HS carrying modifier

Simulation Results for the Inbred Outbred Design

• QTL explaining 10% of phenotypic variance

• 1,500 animals, 500 simulations

• 5 markers per 100 cm detection phase

• Markers at 1 cm interval for fine-mapping

• (We haven’t explored all parameters yet)

Detection Phase

0

2

4

6

8

10

12

14

16

0 20 40 60 80

Log P

Fine-mapping

0

2

4

6

8

10

12

14

16

18

20

0 20 40 60 80 100

Detection

Fine-map

Combined

0

2

4

6

8

10

12

14

16

18

20

0 20 40 60 80 100

Detection

Fine-map

Full

Position Estimates

Summary(2)

• Mapping in a genetically heterogeneous stock of known ancestry can achieve sub-centimorgan resolution

• Theoretically, crosses between inbreds and outbreds can detect and fine-map a genetic effect in one experiment

Acknowledgements

• Chris Talbot • Al Collins• Jeanne Wehner• John DeFries

Wellcome Trust Centre for Human Genetics, Oxford

Institute for Behavioral Genetics, Boulder