9 Natural Selection, Genetic Variation, and Human Diversity
221
signals of selection in metabolic genes are pervasive and striking, it is worth noting
that many other unrelated pathways and processes have been subject to selection as
well.
One important question that genome-wide scans have answered is whether the
targets of the recent selection show evidence of selection in one population or all
populations. Although the number of populations studied so far is disappointingly
small, they usually represent a worldwide, rather than a regional, sample, so any
sharing of candidate selection regions between populations is likely to represent
a worldwide pattern. Several studies have so far observed a significant amount
of overlap between candidate selection regions identified in disparate populations.
One study found evidence of substantial levels of parallel divergence, i.e., sharing
of signals of selection at a gene in phylogenetically independent populations
(Tennessen and Akey 2011). Other studies that do not account for the shared
ancestry between populations see even more evidence of shared targets of selection;
for example, around 30% of targets identified in one study were shared between at
least two worldwide regions (where the world’s populations are clustered into six
world regions) (Pickrell et al. 2009). Although these observations show that a nonnegligible number of candidate selection regions are shared between populations to
some extent, it is also clear that most detectable signatures of positive selection are
geographically restricted. However, the geographic distribution of selected alleles
is not markedly different from randomly selected variants (Coop et al. 2009).
Furthermore, in most cases, the specific allele under selection cannot be identified;
so, although a signal for selection may be seen in multiple populations, it remains
to be seen whether the signal is produced by the same causative allele or not. The
uniqueness of selective forces on different human populations is unsurprising in the
context of human population history but has significant implications for translating
medical applications between populations.
Another issue that genome-wide scans for selection hope to address is whether
the signatures of recent selection we see in the human genome are largely
attributable to alleles of small effect (conferring a small selective advantage) or
large effect (conferring a large selective advantage). For the most part, this question
remains unanswered. The strongest signals of selection are, of course, the easiest
to detect; as a result, candidate selection regions identified so far are biased toward
alleles of large effect (Akey et al. 2002). Since the signature of selection left in
the genome is proportional to the selection coefficient for the advantageous allele,
alleles of small effect are likely to remain outside of our detection ability even
as methods improve. Some studies, however, have attempted to address this issue
by looking for signatures of selection across many genes at once (Hancock et al.
2010b). In these studies, it is possible to detect a very subtle shift in allele frequency
that is pervasive across many genes and, in some cases, correlated with variables
related to environmental selective pressures (Hancock et al. 2010a).
9.4.2.2 Improving Genome-Wide Scans
Most genome-wide scans for selection performed so far have several major weaknesses that limit the applicability of their results. First, with existing methods, only
221
signals of selection in metabolic genes are pervasive and striking, it is worth noting
that many other unrelated pathways and processes have been subject to selection as
well.
One important question that genome-wide scans have answered is whether the
targets of the recent selection show evidence of selection in one population or all
populations. Although the number of populations studied so far is disappointingly
small, they usually represent a worldwide, rather than a regional, sample, so any
sharing of candidate selection regions between populations is likely to represent
a worldwide pattern. Several studies have so far observed a significant amount
of overlap between candidate selection regions identified in disparate populations.
One study found evidence of substantial levels of parallel divergence, i.e., sharing
of signals of selection at a gene in phylogenetically independent populations
(Tennessen and Akey 2011). Other studies that do not account for the shared
ancestry between populations see even more evidence of shared targets of selection;
for example, around 30% of targets identified in one study were shared between at
least two worldwide regions (where the world’s populations are clustered into six
world regions) (Pickrell et al. 2009). Although these observations show that a nonnegligible number of candidate selection regions are shared between populations to
some extent, it is also clear that most detectable signatures of positive selection are
geographically restricted. However, the geographic distribution of selected alleles
is not markedly different from randomly selected variants (Coop et al. 2009).
Furthermore, in most cases, the specific allele under selection cannot be identified;
so, although a signal for selection may be seen in multiple populations, it remains
to be seen whether the signal is produced by the same causative allele or not. The
uniqueness of selective forces on different human populations is unsurprising in the
context of human population history but has significant implications for translating
medical applications between populations.
Another issue that genome-wide scans for selection hope to address is whether
the signatures of recent selection we see in the human genome are largely
attributable to alleles of small effect (conferring a small selective advantage) or
large effect (conferring a large selective advantage). For the most part, this question
remains unanswered. The strongest signals of selection are, of course, the easiest
to detect; as a result, candidate selection regions identified so far are biased toward
alleles of large effect (Akey et al. 2002). Since the signature of selection left in
the genome is proportional to the selection coefficient for the advantageous allele,
alleles of small effect are likely to remain outside of our detection ability even
as methods improve. Some studies, however, have attempted to address this issue
by looking for signatures of selection across many genes at once (Hancock et al.
2010b). In these studies, it is possible to detect a very subtle shift in allele frequency
that is pervasive across many genes and, in some cases, correlated with variables
related to environmental selective pressures (Hancock et al. 2010a).
9.4.2.2 Improving Genome-Wide Scans
Most genome-wide scans for selection performed so far have several major weaknesses that limit the applicability of their results. First, with existing methods, only