4 Types of Natural Selection and Tests of Selection
77
toward rare alleles can be much more pronounced, especially soon before or soon
after the fixation of a beneficial mutation. Unlike beneficial mutations, weakly
deleterious mutations do not create an excess of high-frequency derived alleles.
Unlike hitchhiking, background selection does not drive long haplotypes to high
frequencies (Enard et al. 2014), but it does increase overall linkage disequilibrium
(Zeng and Charlesworth 2011).
4.2
Tests of Selection
As previously described, hitchhiking, balancing selection, and background selection
all have a specific range of effects on patterns of diversity: on overall levels of
diversity, on the SFS, and on the structure of haplotypes. There is considerable
interest in testing the presence of natural selection at specific loci (Akey 2009) in a
genome or genome-wide (Hernandez et al. 2011; Lohmueller et al. 2011; Enard
et al. 2014). Most frequently evolutionary biologists are interested in detecting
local signals of hitchhiking. Balancing selection has received less attention in the
human genome (Andres et al. 2009; Leffler et al. 2013; DeGiorgio et al. 2014).
There has been little interest in detecting background selection at specific loci, and
background selection is usually quantified at a global genome-wide scale, not a local
one (Hernandez et al. 2011; Lohmueller et al. 2011; Enard et al. 2014). Here we
will therefore focus on tests aimed at detecting hitchhiking. However, we will also
mention those cases where tests of positive selection are confounded by background
selection or when a specific test can also be used to detect balancing selection.
We will first describe tests of positive selection based on patterns of genetic
diversity. There are two flavors of these tests: those that use the site frequency
spectrum and those that reply on haplotype patterns. We will then describe one
particular test based on the combination of both genetic diversity and divergence,
the MacDonald–Kreitman test. The tests described here are the most commonly
used: Tajima’s D (Tajima 1989), Fay and Wu’s H (Fay and Wu 2000) and likelihood
ratio-based tests (Kim and Stephan 2002; Nielsen et al. 2005) for the SFS, extended
haplotype homozygosity (EHH) (Sabeti et al. 2002), integrated haplotype score
(iHS) (Voight et al. 2006) and cross-population extended haplotype score (XPEHH) (Sabeti et al. 2007) for haplotype structures, and the McDonald–Kreitman
test (McDonald and Kreitman 1991) for approaches based both on diversity and
divergence. Other tests are often variations of the former tests. We have voluntarily
limited our description to those tests that were most extensively used to analyze
selection in the human genome.
4.2.1 Tests of Selection Based on the Site Frequency Spectrum
4.2.1.1 Tajima’s D
Fumio Tajima proposed his test in 1989 (Tajima 1989), based on the summary
statistic D. Although Tajima’s D has been widely used to detect hitchhiking, Tajima
77
toward rare alleles can be much more pronounced, especially soon before or soon
after the fixation of a beneficial mutation. Unlike beneficial mutations, weakly
deleterious mutations do not create an excess of high-frequency derived alleles.
Unlike hitchhiking, background selection does not drive long haplotypes to high
frequencies (Enard et al. 2014), but it does increase overall linkage disequilibrium
(Zeng and Charlesworth 2011).
4.2
Tests of Selection
As previously described, hitchhiking, balancing selection, and background selection
all have a specific range of effects on patterns of diversity: on overall levels of
diversity, on the SFS, and on the structure of haplotypes. There is considerable
interest in testing the presence of natural selection at specific loci (Akey 2009) in a
genome or genome-wide (Hernandez et al. 2011; Lohmueller et al. 2011; Enard
et al. 2014). Most frequently evolutionary biologists are interested in detecting
local signals of hitchhiking. Balancing selection has received less attention in the
human genome (Andres et al. 2009; Leffler et al. 2013; DeGiorgio et al. 2014).
There has been little interest in detecting background selection at specific loci, and
background selection is usually quantified at a global genome-wide scale, not a local
one (Hernandez et al. 2011; Lohmueller et al. 2011; Enard et al. 2014). Here we
will therefore focus on tests aimed at detecting hitchhiking. However, we will also
mention those cases where tests of positive selection are confounded by background
selection or when a specific test can also be used to detect balancing selection.
We will first describe tests of positive selection based on patterns of genetic
diversity. There are two flavors of these tests: those that use the site frequency
spectrum and those that reply on haplotype patterns. We will then describe one
particular test based on the combination of both genetic diversity and divergence,
the MacDonald–Kreitman test. The tests described here are the most commonly
used: Tajima’s D (Tajima 1989), Fay and Wu’s H (Fay and Wu 2000) and likelihood
ratio-based tests (Kim and Stephan 2002; Nielsen et al. 2005) for the SFS, extended
haplotype homozygosity (EHH) (Sabeti et al. 2002), integrated haplotype score
(iHS) (Voight et al. 2006) and cross-population extended haplotype score (XPEHH) (Sabeti et al. 2007) for haplotype structures, and the McDonald–Kreitman
test (McDonald and Kreitman 1991) for approaches based both on diversity and
divergence. Other tests are often variations of the former tests. We have voluntarily
limited our description to those tests that were most extensively used to analyze
selection in the human genome.
4.2.1 Tests of Selection Based on the Site Frequency Spectrum
4.2.1.1 Tajima’s D
Fumio Tajima proposed his test in 1989 (Tajima 1989), based on the summary
statistic D. Although Tajima’s D has been widely used to detect hitchhiking, Tajima