Finding the Molecular Basis of Quantitative Genetic Variation

1
Finding the Molecular Basis of Quantitative
Genetic Variation
Richard Mott Wellcome Trust Centre for Human
Genetics Oxford UK
2
Genetic Traits

• Quantitative (height, weight)
• Dichotomous (affected/unaffected)
• Factorial (blood group)
• Mendelian - controlled by single gene (cystic
  fibrosis)
• Complex controlled by multiple
  genesenvironment (diabetes, asthma)

3
Molecular Basis of Quantitative Traits
QTL Quantitative Trait Locus
chromosome
genes
4
Molecular Basis ofQuantitative Traits
QTL Quantitative Trait Locus
chromosome
QTG Quantitative Trait Gene
5
Molecular Basis ofQuantitative Traits
QTL Quantitative Trait Locus
chromosome
SNP Single Nucleotide Polymorphism
QTG Quantitative Trait Gene
QTN Quantitative Trait Nucleotide
6
Association Studies

• Compare unrelated individuals from a population
• Phenotypes
• Cases vs Controls
• Quantitative measure
• Genotypes state of genome at multiple variable
  locations (Single Nucleotide Polymorphism SNP)
  in each individual
• Seek correlation between genotype and phenotype

7
Problems with Association Studies

• Population stratification
• Linkage Disequilibrium
• Allele Frequencies
• Multiple loci
• Small Effect Sizes
• Very few Successes

8
Population Stratification

• If the sampling population comprises genetically
  distinct sub-populations with different disease
  prevalences
• Then -
• Any variant that distinguishes the
  sub-populations is likely to show disease
  association

9
Admixture Mapping

• Population is homogeneous but each individuals
  genome is a mosaic of segments from different
  populations
• May be used to map disease loci
• multiple sclerosis susceptibility
• Reich et al 2005, Nature Genetics

10
Linkage Disequilibrium
Mouse
11
Effects of Linkage Disequilibrium

• Correlation between nearby SNPs
• SNPs near to QTN will show association
• Risk of false positive interpretation
• But need only genotype tagging SNPs
• 1 million tagging SNPs will be in LD with 50
  of common variants in the human genome

12
The Common-Disease Common-Variant Hypothesis

• Says
• disease-predisposing variants will exist at
  relatively high frequency (i.e. gt1) in the
  population.
• are ancient alleles occurring on specific
  haplotypes.
• detectable in an case-control study using tagging
  SNPs.
• Alternative hypothesis says
• disease-predisposing alleles are sporadic new
  mutations, perhaps around the same genes, on
  different haplotypes.
• families with history of the same disease owe
  their condition to different mutations events.
• Theoretically detectable with family-based
  strategies which do not assume a common origin
  for the disease alleles, but are harder to detect
  with case-control studies (Pritchard, 2001).

13
Power Depends on

• Disease-predisposing alleles
• Effect Size (Odds Ratio)
• Allele frequency
• Sample Size cases, controls
• Number of tagging SNPs
• To detect an allele with odds ratio of 1.25 and
  with allele frequency gt 1, at 5 Bonferroni
  genome-wide significance and 80 power, we
  require
• 6000 cases, 6000 controls
• 0.5 million tagging SNPs, one of which must be
  in perfect LD with the causative variant
• Hirschorn and Daly 2005

14
WTCCCWellcome Trust Case-Control Consortium

• 2000 cases from each of
• Type I Diabetes
• Type II Diabetes
• rheumatoid arthritis,
• susceptibility to TB
• bipolar depression
• . and others
• 3000 common controls
• 0.675 million SNPs
• 10 billion genotypes
• Data expected mid 2006

15
Mouse Models
16
Map inHuman or Animal Models ?

• Disease studied directly
• Population and environment stratification
• Very many SNPs (1,000,000?) required
• Hard to detect trait loci very large sample
  sizes required to detect loci of small effect
  (5,000-10,000)
• Potentially very high mapping resolution single
  gene
• Very Expensive

• Animal Model required
• Population and environment controlled
• Fewer SNPs required (100-10,000)
• Easy to detect QTL with 500 animals
• Poorer mapping resolution 1Mb (10 genes)
• Relatively inexpensive

17
QTL Mapping in Mice using Inbred Line Crosses

• Genetically Homozygous genome is fixed, breed
  true.
• Standard Inbred Strains available
• Haplotype diversity is controlled far more than
  in human association studies
• QTL detection is very easy
• QTL fine mapping is hard

18
Sizes of Mapped Behavioural QTL in rodents ( of
total phenotypic variance)
19
Physiological QTL
20
Effect sizes of cloned genes
21
QTL detection F2 Intercross
X
A
B
22
QTL mapping F2 Intercross
X
X
A
B
F1
23
QTL mapping F2 Intercross
X
X
A
B
F1
F2
24
QTL mapping F2 Intercross
QTL
1
-1
0
0
0
2
-2
F2
F1
25
QTL mapping F2 Intercross
1
-1
0
0
0
2
-2
F2
F1
26
QTL mapping F2 Intercross
Genotype a skeleton of markers across genome
20cM
0
0
2
-2
F2
27
QTL mapping F2 Intercross
AB AA AB BA
AB BA AB BA
AB BA BA BA
BA BA BA AA
BA BA BA AA
0
0
2
-2
BB BB AB AA
F2
28
QTL mapping F2 Intercross
AB AA AB BA
AB BA AB BA
AB BA BA BA
BA BA BA AA
BA BA BA AA
0
0
2
-2
BB BB AB AA
F2
29
Single Marker Association

• Test of association between genotype and trait at
  each marker position.
• ANOVA
• F2 crosses are
• good for detecting QTL
• bad for fine-mapping
• typical mapping resolution 1/3 chromosome 20-30
  cM

30
Increasing mapping resolution

• Increase number of recombinants
• more animals
• more generations in cross

31
Heterogeneous Stocks

• cross 8 inbred strains for gt10 generations

32
Heterogeneous Stocks

• cross 8 inbred strains for gt10 generations

33
Heterogeneous Stocks

• cross 8 inbred strains for gt10 generations

0.25 cM
34
Mosaic Crosses
G3
GN
F20
inbreeding
mixing
chopping up
HS, AI, outbreds
F2, diallele
RI (RIHS, CC)
35
Analysis of mosaic crosses
chromosome
markers
alleles
1
1
2
1
2
1
1
1
2
2
1
2
2
1
1
1
1
2
1
1
2
1
1
1
1
1
2
1
2
2
1
2
1
1

• Want to predict ancestral strain from genotype
• We know the alleles in the founder strains
• Single marker association lacks power, cant
  distinguish all strains
• Multipoint analysis combine data from
  neighbouring markers

36
Analysis of mosaic crosses
chromosome
markers
alleles
1
1
2
1
2
1
1
1
2
2
1
2
2
1
1
1
1
2
1
1
2
1
1
1
1
1
2
1
2
2
1
2
1
1

• Hidden Markov model HAPPY
• Hidden states ancestral strains
• Observed states genotypes
• Unknown phase of genotypes
• - analyse both chromosomes simultaneously
• Output is probability that a locus is descended
  from a pair of strains
• Mott et al 2000 PNAS

37
Testing for a QTL

• piL(s,t) Prob( animal i is descended from
  strains s,t at locus L)
• piL(s,t) calculated using
• genotype data
• founder strains alleles
• Phenotype is modelled
• yi Ss,t piL(s,t)T(s,t) Covariatesi ei
• Test for no QTL at locus L
• H0 T(s,t) are all same
• ANOVA
• partial F test

38
Example Open Field Avtivity

• Mouse Model for Anxiety

39
(No Transcript)
40
OFA Tracking
41
(No Transcript)
42
multipoint
singlepoint
significance threshold
Talbot et al 1999, Mott et al 2000
43
(No Transcript)
44
Relation Between Marker and Genetic Effect
QTL
Marker 2
Marker 1
No effect observable
Observable effect
45
How Much Mapping Resolution do we need?
46
Mapping Resolution in Mouse QTL experiments

• F2
• 25-50 Mb 250-300 genes
• HS
• 1-5 Mb 10-50 genes
• Need More Resolution

47
Other Outbred Populations

• Commercially available outbreds may contain more
  historical recombination
• Potentially finer mapping resolution
• How to exploit it ?

48
MF1 Outbred Mice MF1
49
Analysis of MF1
50
Single Marker Analysis
51
Unknown progenitors

• Sometime in the 1970s.
• LACA x CF
• MF1

52
MF1 resemble HS

• Sequencing revealed very few new variants in MF1
  compared to HS strains
• Variants present in HS strains also present in MF1

53
MF1 as a mosaic of inbred strains
54
Mapping with 30 generation HS
55
Mapping with MF1 mice
Yalcin et al 2004 Nature Genetics
56
Acknowledgements

• Jonathan Flint
• Binnaz Yalcin
• William Valdar
• Leah Solberg

57
Further Reading

• Mouse
• Flint et al Nature Reviews Genetics 2005
• Human
• Hirschhorn and Daly, Nature Reviews Genetics 2005
• Zondervan and Cardon, Nature Reviews Genetics
  2004