1. Topic #3 Linkage Disequilibrium, Haplotypes & Tagging
University of Wisconsin
Genetic Analysis Workshop
June 2011

2. Overview Fate of a new mutation Linkage Disequilibrium (LD)
Measurement
Indirect association
SNP selection based on LD
Haplotypes
SNP selection by tagging
Practical – SNP selection using Haploview

3. Introduction of a Mutation into a Population
TIME

4. Introduction of a Mutation into a Population
TIME

5. Haplotype Concept
The sequence in this location becomes a signature for the chromosome carrying the mutation
Haplotype – alleles inherited together at linked loci on the same chromosome
haplotype will not be a perfect marker of disease
At the time mutation arose, there may have been other chromosomes with
New mutations
Recombination

7. Indirect Association
Each of the alleles in the haplotype is also expected to be indirectly associated with carrying the mutation.
Indirect association is an association of a marker with phenotype that is non-causal, being based on linkage disequilibrium (LD)

8. Linkage Disequilibrium (LD)
Mendel’s Second Law: alleles at different loci assort independently
Linkage Disequilibrium (LD): population-level association of alleles at linked loci

9. LD – population-level association between linked loci
How LD is Measured
LD – population-level association between linked loci
A locus:
A1 or A2
B locus:
B1 or B2
Let P(A1) = pA1
Let P(B1) = pB1
Let P(A1B1) = pA1B1
D = pA1B1 - pA1pB1 = 0 if independent

10. Common LD Measures D = |d| D’ = |d|/dmax r2 ( D2) = D2/p(1-p)q(1-q)
Preferred measure for population geneticists
Maximum value is bounded by the marginals
D’ = |d|/dmax
D’ varies between 0 and 1
Does not have an easy interpretation and 1.0 is achieved if one off-diagonal is zero
r2 ( D2) = D2/p(1-p)q(1-q)
Has several interpretations:
= squared (phi) correlation so lies in [0,1].
= c2/N
Directly related to power for indirect association

11. Allelic Association Direct Association Indirect Association
Initially it was thought that we could pick the genes and the (single) genetic variant w/i each gene that was relevant for disease
Indirect Association
The existence of LD opens up the possibility of tests by indirect association – we don’t need to actually test the causal variant but rather need only genotype a marker that is in high LD with the causal variant

12. Indirect and Direct Allelic Association
Direct Association
Indirect Association D M1 M2 D M3
Assess relationship of D locus
indirectly by determining whether
markers (Mi) are associated with
disease – Mi don’t need to be
functional
Assess relationship of D locus
to phenotype directly –
expect D to be a functional
polymorphism in a candidate gene

13. Martin, E.R. et al. (2000). SNPing away at complex disease … AJHG 67: 383-394

15. Dawson, E. et al. (2002). A first-generation LD map of 22
Dawson, E. et al. (2002). A first-generation LD map of 22. Nature 418:

16. Population Differences
Weiss, K.M & Clark, A.G. (2002). Trends in Genetics, 18(1):19-24.

17. Recombination Hotspots
Hotspots typically
span 1-2 kb
Kauppi, L., Jeffreys, A. J., & Keeney, S. (2004). Where the crossovers are: Recombination distributions in mammals. Nature Reviews Genetics, 5,

18. Haplotype Blocks

19. Two- and Three-locus Haplotypes
APOE locus and haplotypes
containing APOE
Martin, E.R. et al. (2000). SNPing away at complex disease … AJHG 67:

20. Two- and Three-locus Haplotypes
3-locus haplotype stronger
signal than individual markers
Martin, E.R. et al. (2000). SNPing away at complex disease … AJHG 67:

21. SNP Selection by Tagging
Basic rationale:
The power for a causal SNP in a sample of size N is equivalent to power of tagging SNP in a sample of size N/r2
Tagging SNP selection:
Based on some reference sample (HapMap)
Two overarching strategies
Pairwise tagging
Multimarker tagging
de Bakker, P. I. W., et al. (2005). Efficiency and power in genetic association studies. Nature Genetics, 37(11),

22. Reference Sample: HapMap (www.hapmap.org)
HapMap Phase 1:
SNP Selection Strategy (yield ~ 1 million):
>1 common SNP every 5 kb, total of 1.3 million before QC
MAF > .05
Some priority for non-synonymous cSNPs
Sample: N=270 (269) individuals from 4 populations
30 trios of Europeans from Utah (CEU)
45 unrelated Han Chinese (CHB)
45 unrelated Japanese (JPT)
30 Yoruban trios from Nigeria (YRI)

23. Reference Sample: HapMap (www.hapmap.org)
Phase 2:
2.1 million additional SNPs
Total now averages ~ 1/per kb; >98% of common variants w/i 5kb
Focus still on MAF > .05
Average max r2 of untyped common SNPs to a typed SNP
Population
HapMap I
HapMap II
YRI
.67
.90
CEU
.85
.96
CHB+JPT
.83
.95

24. Reference Sample: HapMap (www.hapmap.org)
Phase 3:
Expand to N=1115 in 11 ancestral groups 2.1 million additional SNPs
* Sample consists of family triples

25. HAPMAP3, Release 2
Region in NCBI B36
COMT
Phase, Release and Build

26. HapMap Genotyped SNPs in COMT

27. Using Haploview to Identify Tagging SNPs for COMT
Download Data from HapMap
Choose HapMap Download, Phase 3, and Release 2
Choose population
Choose chromosome (22) and region (NCBI B36/hg18)
Transcription starts at 18309; I will start at 18304
Transcription ends at 18337; I will end at 18340
Haploview Analysis
Get LD plot
Run Tagger (pairwise)
Force include/exclude

28. COMT LD Plot (D’)

29. COMT LD Plot (r2)

30. COMT Tagging SNPs (15 tag 24 at avg r2 = .996)
Tag SNP bp
Location
MAF
Other SNPs Tagged rs 5’
.37 rs rs
.21 rs
rs737865
Intron #1
.32
rs737866, rs , rs
rs174675
.31
rs933271 rs
.26 rs
.41
rs740601
Intron #2
.45
rs , rs
rs4680
Exon #4
.47
rs4633 rs
Intron #5
.25
rs174696
.18 rs
.11 rs
.17
rs165728
rs165824
.06
rs165815 3’
.12 rs

32. LD Plot Available from SNPInfo(

33. Conclusions
Alleles at linked loci tend to be inherited together, a phenomenon known as linkage disequilibrium (LD)
Because recombination is not uniform, the genome has a “block-like” structure – haplotype
You do not need to have the “causal variant” in your genotyped set if it is adequately tagged
A major strategy for SNP selection is to ensure adequate coverage (r2 > .8) of common genetic variants in a gene, which can be done with Haploview