9 Natural Selection, Genetic Variation, and Human Diversity
225
Krakauer 1973; Nielsen 2005), and hitchhiking and background selection will also
affect differentiation at linked neutral sites. The nature of these unusual patterns
of differentiation depends largely on the type of selection acting there, as well as
whether the selective pressure is acting in the same direction in each population.
To understand the relationship between selection and population differentiation, we
must first identify extremely differentiated (or undifferentiated) regions and then
determine whether this pattern is in fact due to selection.
As discussed above, most targets of selection that have been identified to date are
under selection in a subset of populations. Intuitively, population-specific balancing
selection can make two populations appear to be more differentiated than they
actually are at the locus of interest. The population affected by selection will
maintain several intermediate frequency alleles, while the neutral variation in the
other population will be randomly fixed or lost due to drift. If a locus is under
positive selection in one population but neutral in another, this will also make
the two populations appear to be more differentiated than they actually are (Fig.
9.3a), assuming migration rates are low (Slatkin and Wiehe 1998). The population
undergoing selection will have a different major allele at a very high frequency.
Taken to an extreme, the beneficial allele could be private to the selected population.
In contrast, negative selection in one population, with neutrality in the other, will
make it appear to be less differentiated from another population (Charlesworth et
al. 1997). The population affected by negative selection will be purged of linked
neutral variation, while the other population’s diversity in the region will continue
to be ruled by genetic drift.
If selection is acting in both populations of interest, the situation is more
complicated and will depend on the origin of the selected allele, migration rates
between populations, and geographic factors (Kim and Maruki 2011). In this case,
long-term balancing selection can make the two populations appear to be less
differentiated than they actually are at linked neutral sites (Charlesworth et al.
1997; Schierup et al. 2000). Both populations will be under selection to maintain
ancient alleles, as well as to acquire and maintain many novel alleles. However, the
theoretical models used to make this prediction lack much of the complexity found
in empirical examples of both frequency-dependent and overdominant selection. In
fact, the HLA loci, which comprise the best-known example of balancing selection
in the human genome, exhibit extremely high population differentiation and an
extremely old TMRCA that predates human population divergence (Meyer and
Thomson 2001). This ancient TMRCA indicates very deep splits in the gene tree,
which can appear to be deep population divergence. There are many HLA alleles
that are private to a single population, and most alleles are found at drastically
different frequencies in different populations. Measures of genetic distance between
populations are extremely high at HLA compared to other genomic regions, even
between very closely related populations. The only groups that show reduced
distance measures are extremely isolated populations, which have experienced
drastic founder effects (Vina et al. 2012). The discrepancy between the theoretical
predictions and empirical observations is probably due to the use of simple biallelic
225
Krakauer 1973; Nielsen 2005), and hitchhiking and background selection will also
affect differentiation at linked neutral sites. The nature of these unusual patterns
of differentiation depends largely on the type of selection acting there, as well as
whether the selective pressure is acting in the same direction in each population.
To understand the relationship between selection and population differentiation, we
must first identify extremely differentiated (or undifferentiated) regions and then
determine whether this pattern is in fact due to selection.
As discussed above, most targets of selection that have been identified to date are
under selection in a subset of populations. Intuitively, population-specific balancing
selection can make two populations appear to be more differentiated than they
actually are at the locus of interest. The population affected by selection will
maintain several intermediate frequency alleles, while the neutral variation in the
other population will be randomly fixed or lost due to drift. If a locus is under
positive selection in one population but neutral in another, this will also make
the two populations appear to be more differentiated than they actually are (Fig.
9.3a), assuming migration rates are low (Slatkin and Wiehe 1998). The population
undergoing selection will have a different major allele at a very high frequency.
Taken to an extreme, the beneficial allele could be private to the selected population.
In contrast, negative selection in one population, with neutrality in the other, will
make it appear to be less differentiated from another population (Charlesworth et
al. 1997). The population affected by negative selection will be purged of linked
neutral variation, while the other population’s diversity in the region will continue
to be ruled by genetic drift.
If selection is acting in both populations of interest, the situation is more
complicated and will depend on the origin of the selected allele, migration rates
between populations, and geographic factors (Kim and Maruki 2011). In this case,
long-term balancing selection can make the two populations appear to be less
differentiated than they actually are at linked neutral sites (Charlesworth et al.
1997; Schierup et al. 2000). Both populations will be under selection to maintain
ancient alleles, as well as to acquire and maintain many novel alleles. However, the
theoretical models used to make this prediction lack much of the complexity found
in empirical examples of both frequency-dependent and overdominant selection. In
fact, the HLA loci, which comprise the best-known example of balancing selection
in the human genome, exhibit extremely high population differentiation and an
extremely old TMRCA that predates human population divergence (Meyer and
Thomson 2001). This ancient TMRCA indicates very deep splits in the gene tree,
which can appear to be deep population divergence. There are many HLA alleles
that are private to a single population, and most alleles are found at drastically
different frequencies in different populations. Measures of genetic distance between
populations are extremely high at HLA compared to other genomic regions, even
between very closely related populations. The only groups that show reduced
distance measures are extremely isolated populations, which have experienced
drastic founder effects (Vina et al. 2012). The discrepancy between the theoretical
predictions and empirical observations is probably due to the use of simple biallelic