9 Natural Selection, Genetic Variation, and Human Diversity
223
wide scans. There are many ways in which selection has likely differed from
this simplified model (Fig. 9.3), and it is unclear how robust current neutrality
test statistics are to deviations from it (Przeworski et al. 2005; Pritchard et al.
2010; Hernandez et al. 2011). In particular, models of selection from standing
variation are intuitively well-suited for recent human evolution, which has been
characterized by adaptation to novel environments (Hermisson and Pennings 2005;
Pritchard et al. 2010; Hernandez et al. 2011). In this model, a neutral allele is
drifting in the population when a change in selective pressure makes it beneficial.
The newly beneficial allele also reaches fixation quickly and drags along linked
neutral variants, but because the allele had time to accumulate different haplotype
backgrounds, the signature of the hitchhiking effect is more difficult to detect (Fig.
9.1b–c) (Hermisson and Pennings 2005; Przeworski et al. 2005). All selective sweep
models are also simplified by considering only single-locus effects, despite the fact
that epistatic effects can have a drastic impact on selection (Fig. 9.1d). Detection
methods that employ gene networks or other multi-locus models have already begun
to address this issue (Hancock et al. 2010a, b). While a few genome-wide scans
have examined microsatellite variation for signatures of selection (Payseur et al.
2002; Kayser et al. 2003; Storz et al. 2004), no study to date has used a model of
selection that explicitly accounts for their unique properties (e.g., Fig. 9.1e). Finally,
although candidate gene analyses have revealed tantalizing signatures of selection
at structural variants affecting amylase (Perry et al. 2007) and APOBEC3 (Kidd et
al. 2007), methods to identify structural variants under selection remain elusive.
As selection detection methods are further developed, it will remain imperative
that copy number variants, structural variants, indels, and microsatellites are not
overlooked.
9.5
Selection and Population Differentiation
An outstanding question that has emerged from studies of natural selection in
humans to date is how selection has influenced population differentiation (Barreiro
et al. 2008). Previous interpretations of allele frequency differences within and
between human populations have to a large extent been influenced by the imprecise
and cultural constructions of race, which is of questionable biological meaning
(Barbujani et al. 1997; Bamshad et al. 2004; Vitti et al. 2012). A more nuanced and
precise understanding of worldwide patterns of genetic variation will provide insight
into human evolutionary history and be an important framework for interpreting
geographic patterns in the prevalence and burden of disease. Levels of genetic
divergence between two populations is a complicated mosaic of many factors
including time of population splitting, population sizes, rates of migration and
admixture, and population-specific selective pressures. Therefore, a given level
of genetic divergence between two populations could arise from myriad different
evolutionary histories.
Substrates of recent natural selection are likely to show unusually high or low
levels of differentiation relative to neutrally evolving genomic loci (Lewontin and
223
wide scans. There are many ways in which selection has likely differed from
this simplified model (Fig. 9.3), and it is unclear how robust current neutrality
test statistics are to deviations from it (Przeworski et al. 2005; Pritchard et al.
2010; Hernandez et al. 2011). In particular, models of selection from standing
variation are intuitively well-suited for recent human evolution, which has been
characterized by adaptation to novel environments (Hermisson and Pennings 2005;
Pritchard et al. 2010; Hernandez et al. 2011). In this model, a neutral allele is
drifting in the population when a change in selective pressure makes it beneficial.
The newly beneficial allele also reaches fixation quickly and drags along linked
neutral variants, but because the allele had time to accumulate different haplotype
backgrounds, the signature of the hitchhiking effect is more difficult to detect (Fig.
9.1b–c) (Hermisson and Pennings 2005; Przeworski et al. 2005). All selective sweep
models are also simplified by considering only single-locus effects, despite the fact
that epistatic effects can have a drastic impact on selection (Fig. 9.1d). Detection
methods that employ gene networks or other multi-locus models have already begun
to address this issue (Hancock et al. 2010a, b). While a few genome-wide scans
have examined microsatellite variation for signatures of selection (Payseur et al.
2002; Kayser et al. 2003; Storz et al. 2004), no study to date has used a model of
selection that explicitly accounts for their unique properties (e.g., Fig. 9.1e). Finally,
although candidate gene analyses have revealed tantalizing signatures of selection
at structural variants affecting amylase (Perry et al. 2007) and APOBEC3 (Kidd et
al. 2007), methods to identify structural variants under selection remain elusive.
As selection detection methods are further developed, it will remain imperative
that copy number variants, structural variants, indels, and microsatellites are not
overlooked.
9.5
Selection and Population Differentiation
An outstanding question that has emerged from studies of natural selection in
humans to date is how selection has influenced population differentiation (Barreiro
et al. 2008). Previous interpretations of allele frequency differences within and
between human populations have to a large extent been influenced by the imprecise
and cultural constructions of race, which is of questionable biological meaning
(Barbujani et al. 1997; Bamshad et al. 2004; Vitti et al. 2012). A more nuanced and
precise understanding of worldwide patterns of genetic variation will provide insight
into human evolutionary history and be an important framework for interpreting
geographic patterns in the prevalence and burden of disease. Levels of genetic
divergence between two populations is a complicated mosaic of many factors
including time of population splitting, population sizes, rates of migration and
admixture, and population-specific selective pressures. Therefore, a given level
of genetic divergence between two populations could arise from myriad different
evolutionary histories.
Substrates of recent natural selection are likely to show unusually high or low
levels of differentiation relative to neutrally evolving genomic loci (Lewontin and