Transcript Gen660_Lecture7B_GenomicScans2014

Signatures of Selection
Different types of selection leave behind different signatures on the genome
Negative selection: reduces variation at the affected site(s) but also at
neighboring sites through background selection
Positive selection through recent selective sweep: reduces variation flanking
the selected site (even if neutral) due to hitchhiking
Diversifying selection can increase variation since >1 extreme alleles selected
e.g. selection for diverse viral antigens to evade host immune system
Balancing selection can increase variation by maintaining >1 allele in population
e.g. maintained heterozygosity (sickle cell anemia)
OR
different alleles in different subpopulations due to fluctuating environments
1
Signatures of Selection
Also different methods of looking for these signatures
1. Evolutionary rate within species vs. between species
e.g. Ka/Ks ratio & McDonald-Kreitman tests for coding sequences
HKA and multi-locus HKA tests for non-coding sequences
2. Frequency spectrum: frequency of different alleles in the population
e.g. Tajima’s D … Fay & Wu’s H … Fu & Li’s D*
3. Linkage disequillibrium & Haplotype structure
For all of these tests: compare REAL DATA to
a MODEL of what data should look like under neutral evolution …
can also compare test results at specific loci vs. a scan across the genome
2
Signatures of Selection
Also different methods of looking for these signatures
1. Evolutionary rate within species vs. between species
e.g. Ka/Ks ratio & McDonald-Kreitman tests for coding sequences
HKA and multi-locus HKA tests for non-coding sequences
2. Frequency spectrum: frequency of different alleles in the population
e.g. Tajima’s D … Fay & Wu’s H … Fu & Li’s D*
3. Linkage disequillibrium & Haplotype structure
For all of these tests: compare REAL DATA to
a MODEL of what data should look like under neutral evolution …
can also compare test results at specific loci vs. a scan across the genome
3
Methods based on the Allele Frequency Spectrum
1. For each ‘derived’ (=non-ancetsral) allele at a given locus, calculate the frequency.
Some alleles will be at high frequencies in the population,
some at low frequencies (i.e. very uncommon)
2.
Make a histogram of the % of alleles with different frequencies
looking for an excess of rare alleles or of common alleles
4
From Nielsen Nat Rev Gen 2005 review
Methods based on the Allele Frequency Spectrum
Tajima’s D (F. Tajima, 1989): takes the # of segregating sites within species (S)
and also the average # difference between each pair of sequences ()
S=3
 = (2 + 2 + 1 + 2) + (2 + 1 + 0) + (1 + 2) +(1) = 1.4
10 pairwise comparisons
avg. # difs between
each pair of sequences
5
Tajima’s D compares S and  to estimate the proportion of low/high-frequency alleles
Methods based on the Allele Frequency Spectrum
Tajima’s D (F. Tajima, 1989): takes the # of segregating sites within species (S)
and also the average # difference between each pair of sequences ()
S versus  reflects on allele frequency
Multiple ways to calculate q
q =   q = S/a
Negative Tajima’s D = excess of low-frequency alleles (= reduced variation)
( < S/a)
Indicates positive selection, OR recent deleterious alleles, OR population expansion**
Positive Tajima’s D = excess of intermediate-frequency alleles
( > S/a)
(low amounts of both high- and low-frequency alleles)
Indicates balancing selection OR partial sweep OR population bottleneck**
How can you get a p-value? Difficult to estimate - best to compare across loci
6
Empirical model for significance of Tajima’s D
Sliding window across a locus
From Nielsen Nat Rev Gen 2005 review
OR
Compare to several other loci
From Will et al. PLoS Genetics 2010
7
Genome-wide scans of FST
FST is a measure of population subdivision:
the proportion of the total genetic variance T contained in a subpopulation S
relative to the total genetic variance in the species
FST =
T - S
T
Where  = average # pairwise nucleotide differences per site
If S = T (i.e amount of variation in the subpopulation is same as total population)
FST = 0 … NO population subdivision
If there’s variation in the total sample, but NO variation within each subpopulation
S =  FST = 1 … COMPLETE differentiation between subpopulations
8
FST = 1: very strong population
subdivisions … may be
little gene flow between
those populations
9
Genome-wide scans of FST
Difficult to interpret what a given FST means (FST = 0.15 means ???)
But, can use variation in FST across the genome to look for evidence
of partial selective sweeps in specific sub-populations:
i.e. little gene flow at specific loci only
10
From Akey et al. 2002: FST across each human chromosome
LD & Haplotype Structure
Linkage equillibrium: when segregation of two different alleles is independent of one another
Linkage disequillibrium (LD): segregation of two alleles are NOT random
- two SNPs in close proximity are linked physically
- can measure the distance over which their association breaks down
LD break-down depends on generation time and recombination rate
SNPs very close together will take
many generations to get separated
12
LD & Haplotype Structure
Linkage equillibrium: when segregation of two different alleles is independent of one another
Linkage disequillibrium (LD): segregation of two alleles are NOT random
- two SNPs in close proximity are linked physically
- can measure the distance over which their association breaks down
Haplotype: block of linked SNPs
Haplotype 1 at Locus A
Haplotype 2 at Locus A
Haplotype 3 at Locus A
13
LD & Haplotype Structure
Remember that a recent selective sweep can reduce variation flanking
the advantageous site.
The strength of selection and time since sweep affects the degree and length of reduced variation.
This effectively
creates an unusually
long haplotype
(compared to others
in the genome)
14
EHH: Extended Haplotype Homozygosity test
for RECENT positive selection
Recent positive selection through partial selective sweep:
* extended haplotype length
* high frequency in subpopulation
must account for regional differences in recombination rates
Yoruban
Beni
African
Shona
European
Asian
15
EHH: Extended Haplotype Homozygosity test
for RECENT positive selection
EHH = % of individuals sharing CORE haplotype that remain identical
out to a distance of x
Defined Core Haplotype
16
EHH: Extended Haplotype Homozygosity test
for RECENT positive selection
Relative EHH: normalize EHH for one haplotype to EHH of all others at that locus
internally controls for locus-specific effects
African haplotype
17
EHH: Extended Haplotype Homozygosity (& other methods) test
for RECENT positive selection
Related test from Jonathan Pritchard: iHS test
Benefits of EHH & iHS scans:
* Don’t have to know populations a priori … define by haplotypes
* More sensitive than traditional tests for selection
Remaining challenges:
* Often have no idea WHY - how to link to phenotypes of interest?
Stinchcombe & Hoekstra review: combining scans with QTL mapping
* Often unclear what SNP was selected for … identifies huge regions
18
Science. February 12, 2010
CMS incorporates results of 5 different tests:
FST
iHS & XP-EHH
DAF (looking at derived allele frequencies)
iHH (looking at absolute haplotype length)
19
CMS outperforms single tests in simulated data
20