some current issues in qtl identification lon cardon wellcome trust centre for human genetics...

Some current issues in QTL identification

Lon CardonWellcome Trust Centre for Human Genetics

University of Oxford

Acknowledgements: Goncalo AbecasisStacey ChernyTwin course faculty

Positional Cloning

LO

D

Sib pairs Chromosome Region Association Study

Genetics

GenomicsPhysical Mapping/Sequencing

Candidate Gene Selection/Polymorphism Detection

Mutation Characterization/Functional Annotation

Inflammatory Bowel Disease Genome Screen

Hampe et al., Am J Hum Genet, 64:808-816, 1999

Inflammatory Bowel Disease Genome Screen

Hampe et al., Am J Hum Genet, 64:808-816, 1999

Susceptibility locus mapped for Crohn’s Disease

Genome Screens for Linkage in Sib-pairs

1997/98- Diabetes (IDDM + NIDDM)- Asthma/atopy- Osteoporosis- Obesity- Multiple Sclerosis- Rheumatoid arthritis- Systemic lupus erythematosus- Ankylosing spondylitis- Epilepsy- Inflammatory Bowel Disease- Celiac Disease- Psychiatric Disorders (incl. Scz, bipolar)- Behavioral traits (incl. Personality, panic)- others missed...

1999- NIDDM- Asthma/atopy- Psoriasis- Inflammatory Bowel Disease- Osteoporosis/Bone Mineral Density- Obesity- Epilepsy- Thyroid disease- Pre-eclampsia- Blood pressure- Psychiatric disorders (incl. Scz, bipolar)- Behavioral traits (incl. smoking, alcoholism,

autism)- Familial combined hyperlipidemia- Tourette syndrome- Systemic lupus erythematosus- others missed…

Human QTL Linkage Gene Identification Successes

0Well, at least < 5

Why so few successes in human QTL mapping?

Many valid reasons proposed:• Phenotypic complexity (not measured well)• Genetic complexity (many genes of small effect, GxE, epistasis)• Genotype error• Sampling design• Statistical methods• ….

Most linkage studies have been under-powered (and over-hyped)

QTL Mapping has very low power !1000 sibs, no parents: markers every 10 cM, each marker H=0.8

QTLh2=0.33

Kruglyak L, Lander ES. (1995). Am J Hum Genet 57: 439-454

Increasing power to detect linkage in sib-pairs

• Phenotypic selection– Carey & Williamson, 1991, AJHG

– Eaves & Meyer, 1994, Behav Genet

– Cardon & Fulker, 1994, AJHG

– Risch & Zhang, 1996, AJHG

Equivalent full sample N for 200 selected pairs from 10,000 (QTL allele freq = .2)

Concordant Discordant Combined

Additive 1400 3300 5000

Recessive 6000 3100 9500

Dominant 1400 3100 4400

1 2 3 4 5 6 7 8 9 102

46

810100

150

200

250

300

350

Decile ranking - Sib 1

Sib 2

Info

rma

tio

n s

core

Information Score for Additive Gene Action (p=0.5)

Linkage Analysis of QTLs-Summary-

• Spotted history. Few, if any, bona fide successes• Power has been large problem

• Of the few replicated loci, most have used some form of selection• EDAC, other selection schemes from large cohorts now underway• Genome-scans coming soon

Promising beginning for QTL linkage mapping

Positional Cloning

LO

D

Sib pairs Chromosome Region Association Study

Genetics

GenomicsPhysical Mapping/Sequencing

Candidate Gene Selection/Polymorphism Detection

Mutation Characterization/Functional Annotation

Association Analysis

• Simple genetic basis

Short unit of resemblancePopulation-specific

• One of easiest genetic study

designs

Correlate allele frequencies with traits/diseasesAt core of monogenic & oligo/polygenic trait models

• Widely used in past 20 years

HLA, candidate genes, pharmacogenetics, positional cloning

Angiotensin-1 Converting Enzyme

Keavney et al. (1999) Hum Mol Gen, 7:1745-1751

Evidence for Linkage

0

5

10

LO

D

A-5466C A-240T T1237C I/D 4656(CT)3/2

T-5991C T-3892C T-93C G2215A G2350A

Results of ACE analysis using VC association model

A-5466C A-240T T1237C I/D 4656(CT)3/2

T-5991C T-3892C T-93C G2215A G2350A0

5

10

15

LOD

for Linkage for Association

Alzheimers and ApoE4

Roses, Nature 2000

Association Resolution by Position

Roses, Nature 2000

Decay of Linkage Disequilibrium in a Small Set of Genes

Toward a linkage disequilibrium map of the human genome

• > 10 year ago, emphasis mainly on theory - LD measures, decay, population comparisons, …

• 1989: 1st use of LD for disease mapping: Cystic Fibrosis

• Recent years, gene-based haplotypes used widely for monogenic mapping

• Last 2 years: larger scale assessment of common alleles in reference populations

LD/haplotype map objective: find regions of high and low ancestral conservation to clarify signal/noise in allelic association studies

History of LD studies in humans:

Haplotype Map: Data/Interpretations

Distribution of pairwise LD ‘average extent of LD’

LD differences in genes

Eaves et al, Nat Genet 2000 Taillon-Miller et al, Nat Genet 2000

Stephens et al, Science 2001

Reich et al, Nature 2001

Johnson et al, Nat Genet 2001 Abecasis et al, AJHG 2001

Haplotype Map: Data/Interpretations

Local patterns of LD … Conserved haplotype segments ... ‘Blocks’

5q31. Daly et al, Nat Genet 2001

MHC class II. Jeffreys et al, Nat Genet 2001

Chr21. Patil et al, Science 2001

Current Status: Data/Interpretations

• How to define ‘useful’ LD is still unclear

• Easier to focus on pairwise LD rather than haplotypes. Is this efficient?

• For common alleles, D’ measure, LD extends ~ 50-60 kb on averageFor rare alleles, ?

• There is great variability in regional patterns of LDExplanations, predictors yet unknown

• Haplotype blocks are detectable and present broadly

• Size of blocks? How best to define them? Utility of htSNPs?

Human Genome Haplotype Map

1. NIH/TSC/Wellcome Trust funded international collaboration (likely)- follow-on from human sequencing project & SNP consortium

2. Hierarchical strategy- ‘sparse-map’ then more fine- Initially use available SNPs

3. Multiple populations- some family-based, most likely to be unrelateds

4. Aim is to catalog regions of high LD down to very fine-scale (ie., find big and small blocks)

Human Chromosome 22• First human chromosome to be “fully” sequenced

• Extensive knowledge of genomic landscape

• Abundance of SNPs and other variants/bp

~34.5 Mb on q-arm; p-arm mostly structural RNA; 679 genes on qDunham et al, Nature, 1999

Samples

• 7 x 3 generation CEPH families– 77 Individuals– 59 founder chromosomes– 1505 SNPs successfully genotyped

• 90 Unrelated Caucasian Individuals– 1286 SNPs genotyped (1261 overlapping with CEPHs)

• 51 Unrelated Estonian Individuals– 908 SNPs genotyped (594 overlapping with CEPHs)

N = 1505 markers. Median spacing = 15.07kb. 4 gaps > 200 kb. Smallest = 12 bp; largest = 293 kb.

Marker spacing

0

100

200

300

400

500

600

< 5k

b5-

1011

-20

21-3

0

31-4

0

41-5

0

51-6

0

61-7

0

71-8

0

81-9

0

91-1

00

101-

110

111-

120

121-

130

131-

140

> 15

0kb

Spacing bin

Co

un

t

N=1505

Allele frequencies on Chromosome 22Ceph founders

0

0.05

0.1

0.15

0.2

0.25

< 0.10 .11-.20 .21-.30 .31-.40 < 0.50

Category

Fre

qu

ency

0.00

0.20

0.40

0.60

0.80

1.00

0 200 400 600 800 1000

Physical Distance (kb)

D'

0.00

0.20

0.40

0.60

0.80

1.00

0 200 400 600 800 1000

Physical Distance (kb)

r2

D’

r2

Variability in Pairwise LD

Decay of LD on chromosome 22Means in CEPHs, Unrelateds, Combined & Estonian Samples

Representing LD along a chromosome

Following several trends in genetics, genotyping technology outpaced ability to analyze LD information…

How to characterize regions of ‘interesting’ linkage disequilibrium?

1. Simply examine average levels across region/chromosome?2. Fit models to data, look at expectations & specific predictions3. Consider ‘interesting’ LD tracts as long runs of LD – borrow from

extant statistical approaches4. Look for ‘blocks’ of LD in the genome

LD Along Chromosome 22

0.00

0.25

0.50

0.75

1.00

0 5 10 15 20 25 30

D'

0

200

400

600

0 5 10 15 20 25 30

Position (Mb)

Pre

dic

ted

Hal

f-L

ife

(kb

)Average D’

D’ Half-Life

Disequilibrium Fingerprint

Plus 3 individual blocks:Position SNPs Haplos Length4.6-4.8 M 11 6 231 kb8.2-8.4 M 8 4 264 kb34.3 M 11 3 82 kb

Chromosome 22 Haplotype Blocks

Chr22 High LD: 22-27 Mb

Chr22 Low LD: 27-32 Mb

Recombination Pattern on Chromosome 22

1 Mb/cM

Microsatellite distance

0

10

20

30

40

50

60

0 5 10 15 20 25 30 35

Sequence Position (Mb)

cM

1 Mb/cM

Microsatellite distance

0

10

20

30

40

50

60

0 5 10 15 20 25 30 35

Sequence Position (Mb)

cM

GeneDensity

Recombination and Gene Density on Chromosome 22

Linkage Disequilibrium Map of Chromosome 22 - Summary -

• LD ‘half-length’ ~ 50 kb, but depends on measure & what is “useful” LD

• Family & unrelated samples yield consistent patterns

• Different analytical tools provide complementary views of long blocks

• 15% chromosome 22 in long LD blocks in these samples (40% in shorter blocks) Why? Selection, selective sweeps? Chromosome structure? Popln age?

• LD correlated with gene-density, GC content and related repeats.Gene/GC correlations almost entirely collinear with genetic distance.

LD patterns can immediately assist positional association studies:

Prioritise candidate regions.Use extant genetic maps and simple repeat structures in design & power.

Mapping QTLs in families:Summary

• Linkage and association studies follow directly from fundamental biometrical principles.

• Linkage studies of complex traits can work: All principles of this course apply

- power, study design, careful phenotype selection/modelling, comparison of statistical models

• New information about LD patterns should facilitate association studies

- help form a priori hypotheses and guide replication.

16th Annual Course on Methodology for Twins and FamiliesAdvanced workshop: Boulder, Colorado, March 2003

Monday, 5 March 2001

Eaves 9:00-10:30 Introduction: Cause of human variation

Amos & Heath 11:00-12:00 Basic Statistics: Likelihood models

Lessem 12:00-12:30 Introduction: Computer System P

Eaves & Sham 13:30-15:00 Genetic Theory

Neale, Martin & Boomsma 15:30-17:00 MX practical P

Tuesday, 6 March 2001

Sham 9:00-10:30 Linkage: Basic Principles

Abecasis, Cherny & Cardon 11:00-12:30 IBD estimation: Theory and Practice P

Martin & Maes 13:30-15:00 QTL Linkage Analysis in Sibships P

Eaves 15:30-17:00 Introduction to Bayesian Methods P

Wednesday, 7 March 2001

Neale & Heath 9:00-10:30 Linkage on Selected Samples P

Purcell & Sham 11:00-12:30 Power Calculation in Linkage Analysis P

Boomsma & van Baal 13:30-15:00 Multivariate Applications P

Purcell & Sham 15:30-17:00 Epistasis/Multi-locus modelling P

Thursday, 8 March 2001

Rice & Heath 9:00-10:30 Association Study Principles P

Cherny & Abecasis 11:00-12:30 Family Based Association Studies P

van den Oord 13:30-15:00 Population Stratification and General Association

P

Sham & Abecasis 15:30-17:00 Power for Association Analysis P

Friday, 9 March 2001

Cardon & Sham 9:00-10:30 Bioinformatics and Genome Patterns of Disequilibrium

Rice 11:00-12:30 Multiple Testing: Power and Type I Error

Flint 13:30-15:00 Animal models of complex traits

Cherny, Purcell & Abecasis 15:30-17:00 General computational issues P

http://ibgwww.colorado.edu/twins2001/schedule.html