Linkage Disequilibrium by HC111213171416


									                                      Slide 1

   Linkage Disequilibrium

         Joe Mychaleckyj
Center for Public Health Genomics

                    Joe Mychaleckyj
                                           Slide 2

Today we’ll cover…

•   Haplotypes
•   Linkage Disequilibrium
•   Visualizing LD
•   HapMap

                         Joe Mychaleckyj
                                                                                        Slide 3


                                       Principles of Population Genetics,
                                       Fourth Edition (Hardcover) by Daniel L. Hartl,
                                       Andrew G. Clark (Author)

                                       Genetic Data Analysis II Bruce S Weir
         QuickTime™ an d a
TIFF (Uncompressed) decompressor
   are need ed to see this picture .


                                                          Joe Mychaleckyj
                                            Slide 4


       Statistical Genetics: Gene Mapping Through
       Linkage and Association Eds Benjamin M.
       Neale, Manuel A.R. Ferreira,
       Sarah E. Medland, Danielle Posthuma

                         Joe Mychaleckyj
                                                                         Slide 5

                   SNP1                 SNP2                SNP3
                   [A / T]              [C / G]             [A / G]
                      A                    C                   G
                      A                    C                   A
                      T                    G                   G
Haplotype: specific combination of alleles occurring (cis) on the same
chromosome (segment of chromosome)

N SNPs - How many Haplotypes are possible ?

          2N (ie very large diversity possible)

                                          Joe Mychaleckyj
                                            Slide 6


• Haplotype: Specific combination
  (phasing) of alleles occurring (cis) on
  the same chromosomal segment
• Linkage/Linked Markers: Physical co-
  location of markers on the same
• Diplotype: Haplogenotype ie pair of
  phased haplotypes one maternally, one
  paternally inherited

                        Joe Mychaleckyj
                                                             Slide 7

                 SNP1 [ A / a ]           SNP2 [ B / b ]

Major Allele Freq:     p(A)                           p(B)
Minor Allele Freq:     p(a)                           p(b)
 Independently segregating SNPs:
 Haplotype Frequency p(ab) = p(a) x p(b)

 (How many haplotypes in total ?)

 Haplotype Frequency p(ab)≠ p(a) x p(b)

                                    Joe Mychaleckyj
                                          Slide 8

Linkage Disequilibrium

• Non-random assortment of alleles at 2
  (or more) loci
• The closer the markers, the stronger
  the LD since recombination will have
  occurred at a low rate
• Markers co-segregate within and
  between families

                        Joe Mychaleckyj
                                                             Slide 9

                                                   Not a Punnett
            SNP2 Allele                               Square!
Allele      B          b
  A      p(A)p(B)             p(A)p(b) p(A)
  a      p(a)p(B)             p(a)p(b) p(a)

                p(B)   p(b)
 p(A)p(B)+p(a)p(B)=p(B){ p(A)+p(a)} = p(B)

                                 Joe Mychaleckyj
                                                             Slide 10

                 SNP1 [ A / a ]           SNP2 [ B / b ]

Major Allele Freq:       p(A)                         p(B)
Minor Allele Freq:       p(a)                         p(b)

 Haplotype Frequency p(ab) = p(a) p(b) + D
 (sign of D is generally arbitrary, unless comparing D values
 between populations or studies)
 D: Lewontin’s LD Parameter (Lewontin 1960)

                                    Joe Mychaleckyj
                                                   Slide 11


          SNP2 Allele
Allele     B          b
  A      p(A)p(B)+D         p(A)p(b)-D p(A)
  a      p(a)p(B)-D         p(a)p(b)+D p(a)

             p(B)         p(b)

 p(A)p(B)+D + p(a)p(B)-D =p(B){ p(A)+p(a)} =
                                 Joe Mychaleckyj
                                                      Slide 12

     b         B

a   0.16      0.04    p(a)=0.20   What is the LD ?
A   0.14      0.66    p(B)=0.80
                                  p(ab) ≠ p(a) p(b)
    p(b)=0.30 p(B)=0.70

                                  p(ab) = p(a) p(b) + D

    0.16 = 0.2 x 0.3 + D
    D = 0.1
    Since p(ab) = p(a)p(b)+ D
    +D was used and D is +ve here, but arbitrary
    eg can relabel alleles A,B as minor
                                  Joe Mychaleckyj
                                                                Slide 13

Range of D values (-ve to +ve)

D has a minimum and maximum value that depends on the allele
frequencies of the markers
Since haplotype frequencies cannot be -ve
p(aB) = p(a)p(B) - D ≥ 0                D ≤ p(a)p(B)
p(Ab) = p(A)p(b) - D ≥ 0                D ≤ p(A)p(b)
These cannot both be true, so D ≤ min( p(a)p(B), p(A)p(b) )
p(ab) = p(a)p(b) + D ≥ 0                D ≥ -p(a)p(b)
p(AB) = p(A)p(B) + D ≥ 0                D ≥ -p(A)p(B)
These cannot both be true, so D ≥ max( -p(a)p(b), -p(A)p(B) )

* Similar equations if we had defined p(ab) = p(a)p(b) - D

                                        Joe Mychaleckyj
                                            Slide 14

Limits of D LD Parameter

 Limits of D are a function of allele

 Standardize D by rescaling to a
 proportion of its maximal value for the
 given allele frequencies (D')
          D’ =    D

                          Joe Mychaleckyj
                                                Slide 15

D’ (Lewontin, 1964)
D’ = D / Dmax
Dmax = min (p(A)p(B), p(a)p(b))       D<0
Dmax = min (p(A)p(b), p(a)p(B))       D>0
Again, sign of D’ depends on definition

D’ = 1 or -1 if one of p(A)p(B), p(A)p(b),
  p(a)p(B), p(a)p(b) = 0
= Complete LD (ie only 3 haplotypes seen)
D’=1 or -1 suggests that no recombination has
  taken place between markers
Beware rare markers - may not have enough
  power/sample size to detect 4th haplotype
                            Joe Mychaleckyj
                                                                        Slide 16

D’ Interpretation

        b       B                         b          B

  a   0.06      0.14   p(a)=0.20    a   0.2           0        p(a)=0.20

  A   0.24      0.56   p(A)=0.80    A   0.1           0.7      P(A)=0.80

      p(b)=0.30 p(B)=0.70               p(b)=0.30 p(B)=0.70

D=0 ; Dmax undefined               D=Dmax =0.14 ; D’ = +1

                                           p(a) = 0.2       p(b)= 0.3
  D’=1 (perfect LD using D’ measure
   - No recombination between marker
   - Only 3 haplotypes are seen

                                        Joe Mychaleckyj
                                                            Slide 17

Creation of LD

• Easiest to understand when markers are
  physically linked
• Creation of LD
  –   Mutation
  –   Founder effect
  –   Admixture
  –   Inbreeding / non-random mating
  –   Selection
  –   Population bottleneck or stratification
  –   Epistatic interaction
• LD can occur between unlinked markers
• Gametic phase disequilibrium is a more
  general term
                                          Joe Mychaleckyj
                                                       Slide 18

SNP1                 SNP1      SNP2

                      A          B        n=3 haplotypes

    n=2 haplotypes
                      A          b

a                     a          B

            SNP1     SNP2

              A        B

              A        b    n=4 haplotypes

              a       B

              a       b
                            Joe Mychaleckyj
                                          Slide 19

Destruction of LD

• Main force is recombination
• Gene conversion may also act at short
  distances (~ 100-1,000 bases)
• LD decays over time (generations of

                        Joe Mychaleckyj
                                                           Slide 20

   SNP1       SNP2        Probability Recombination
                          occurs = θ
                          Probability Recombination
                          does not occur = 1-θ

Initial LD between SNP1 - SNP2: D0

After 1 generation

Preservation of LD:
       D1 = D0(1-θ)
                             NB: Overly simple model -
After t generations:         does not account for allele
                             frequency drift over time
        Dt = D0 (1- θ)t

                            Joe Mychaleckyj
                                                         Slide 21

                          Dt = D0 (1-θ)t

         QuickTime™ an d a
TIFF (Uncompressed) decompressor
   are need ed to see this picture .

                                       Joe Mychaleckyj
                                                    Slide 22

r2 LD Parameter (Hill & Robertson, 1968)
   r   2   =         D2

 • Squared correlation coefficient varies 0 - 1
 • Frequency dependent
 • Better LD measure for allele correlation
   between markers - predictive power of SNP1
   alleles for those at SNP2
 • Used extensively in disease gene or
   phenotype mapping through association

                                  Joe Mychaleckyj
                                                                                 Slide 23

r2 Interpretation

           b        B                               b               B

    a    0.06        0.14 p(a)=0.20         a     0.2               0     p(a)=0.20

    A    0.24        0.56 p(A)=0.80         A     0.1               0.7   p(A)=0.80

         p(b)=0.30 p(B)=0.70                      p(b)=0.30 p(B)=0.70

D=0 ; Dmax undefined                      D=Dmax =0.14 ; D’ = +1

r2 = 0                                    r2 = 0.14/0.24 = 0.58

                                                p(a) = 0.2    p(b) = 0.3
r2 ≠ 1 Correlation is not perfect, even
though D’ = 1
r2 = 1 if D’ = 1 and p(a) = p(b) = 0.3

                                                  Joe Mychaleckyj
                                                                Slide 24

r2 Interpretation
                                      p(a) = 0.3   p(b) = 0.3
Only 2 haplotypes:

r2 = 1 Correlation is perfect
D’ =1 (less than 4 haplotypes)
p(a) = p(b) (= 0.3 in this example)

   • r2=1 when there is perfect correlation between
     markers and one genotype predicts the other exactly
       – Only 2 haplotypes present
   • D’ = 1 ≠> r2 = 1
   • No recombination AND markers must have identical
     allele frequency
       – SNPs are of similar age
   • Corollary
      – Low r2 values do not necessarily = high recombination
      – Discrepant allele frequencies Joe Mychaleckyj
                                                   Slide 25

Common Measures of Linkage Disequilibrium

   -1                 D’                       1

    0                 r2                       1

     Other LD Measures exist, less
     common usage
                             Joe Mychaleckyj
                                  Slide 26

Visualizing LD metrics

                Joe Mychaleckyj
                                                   Slide 27

                    1 2     3    4      5      6
| D’ |
0.8       SNP2
0.6       SNP3

 Not usually worried about sign of D’
                                 Joe Mychaleckyj
                  Slide 28

Joe Mychaleckyj
                                         Slide 29

Haploview: TCN2 (r2)

                       Joe Mychaleckyj
                                              Slide 30


Launched October 2002

                            Joe Mychaleckyj
                                                                 Slide 31

International HapMap
• Initiated Oct 2002
• Collaboration of scientists worldwide
• Goal: describe common patterns of human
  DNA sequence variation
• Identify LD and haplotype distributions
• Populations of different ancestry (European,
  African, Asian)
   – Identify common haplotypes and population-specific
• Has had major impact on:
   – Understanding of human popualtion history as reflected in
     genetic diversity and similarity
   – Design and analysis of genetic association studies

                                      Joe Mychaleckyj
                                                       Slide 32

HapMap samples

• 90 Yoruba individuals (30 parent-parent-offspring
  trios) from Ibadan, Nigeria (YRI)
• 90 individuals (30 trios) of European descent from
  Utah (CEU)
• 45 Han Chinese individuals from Beijing (CHB)
• 44 Japanese individuals from Tokyo (JPT)

                                 Joe Mychaleckyj
                                                          Slide 33

Project feasible
because of:
• The availability of the human genome sequence
• Databases of common SNPs (subsequently enriched by
  HapMap) from which genotyping assays could be
• Development of inexpensive, accurate technologies for
  highthroughput SNP genotyping
• Web-based tools for storing and sharing data
• Frameworks to address associated ethical and cultural

                                 Joe Mychaleckyj
                                                        Slide 34

HapMap goals

• Define patterns of genetic variation across human
• Guide selection of SNPs efficiently to “tag” common
• Public release of all data (assays, genotypes)
• Phase I: 1.3 M markers in 269 people
   1 SNP/5kb (1.3M markers)
   Minor allele frequency (MAF) >5%
• Phase II: +2.8 M markers in 270 people

                                  Joe Mychaleckyj
                                           Slide 35
                         Joe Mychaleckyj
                  Slide 36

Joe Mychaleckyj
                  Slide 37

Joe Mychaleckyj
                                                                      Slide 38

HapMap publications

•   The International HapMap Consortium. A Haplotype Map of the
    Human Genome.
    Nature 437, 1299-1320. 2005.

•   The International HapMap Consortium. The International
    HapMap Project.
    Nature 426, 789-796. 2003.

•   The International HapMap Consortium. Integrating Ethics and
    Science in the International HapMap Project.
    Nature Reviews Genetics 5, 467 -475. 2004.

•   Thorisson, G.A., Smith, A.V., Krishnan, L., and Stein, L.D. The
    International HapMap Project Web site.
    Genome Research,15:1591-1593. 2005.

                                            Joe Mychaleckyj
                                                Slide 39

ENCODE project

• Aim: To compare the genome-wide resource
  to a more complete database of common
  variation—one in which all common SNPs
  and many rarer ones have been discovered
  and tested
• Selected a representative collection of ten
  regions, each 500 kb in length
• Each 500-kb region was sequenced in 48
  individuals, and all SNPs in these regions
  (discovered or in dbSNP) were genotyped in
  the complete set of 269 DNA samples

                           Joe Mychaleckyj
                            Slide 40

      Comparison of linkage
      disequilibrium and
      recombination for two ENCODE

       Nature 437, 1299-1320. 2005

Joe Mychaleckyj
                                Slide 41

LD in Human Populations

              Joe Mychaleckyj
                                                  Slide 42

Haplotype Blocks

N SNPs = 2N Haplotypes possible, ie very large
diversity possible
But: we do not see the full extent of haplotype
diversity in human populations
      Extensive LD especially at short distances eg
Haplotypes are broken into blocks of markers with
high mutual LD separated by recombination hotspots
Non-uniform LD across genome

                                Joe Mychaleckyj
                                                                  Slide 43

   Haplotype Blocks

  Haplotype blocks: at least 80% of observed haplotypes
  with frequency >= 5% could be grouped into common

Whole Genome Patterns of Common DNA Variation
in Three Human Populations, Science 2005, Hinds et al.
                                                Joe Mychaleckyj
                                            Slide 44

Length of LD spans


                     Joe Mychaleckyj
                                                              Slide 45

Example: Large block of LD on chromosome 17
Cluster of common (frequent SNPs In high LD)
518 SNPs, spanning 800 kb
25% in EUR, 9% in AFR, missing in CHN
  Microtubule-associated protein tau
  Mutations associated with a variety of
  neurodegeneartive disorders
  Gene coding for a protease similar to
  Mutations result in Alzheimer’s disease
  Gene for corticotropin-releasing hormone
     • Immune, endocrine, autonomic, behavioral response to
                                     Joe Mychaleckyj
                                          Slide 46

Chromosome 17 LD Region

             Prevalent inversion in EUR
             human population

                     Joe Mychaleckyj

To top