Bamshad, M. & Wooding, S.P. Signatures of natural selection in the human genome. Nat. Rev. Genet. 4, 99-111

Abstract

During their dispersal from Africa, our ancestors were exposed to new environments and diseases. Those who were better adapted to local conditions passed on their genes, including those conferring these benefits, with greater frequency. This process of natural selection left signatures in our genome that can be used to identify genes that might underlie variation in disease resistance or drug metabolism. These signatures are, however, confounded by population history and by variation in local recombination rates. Although this complexity makes finding adaptive polymorphisms a challenge, recent discoveries are instructing us how and where to look for the signatures of selection.

Full text

© 2003
Nature
Publishing
Group
Humans differ from each other in many ways, ranging
from their physical appearance,behaviour and suscepti-
bility to disease, to their likelihood of experiencing an
adverse drug reaction.Although part of this phenotypic
variability results from differences in environmental
exposures or chance,it is clear that gene variants are also
responsible.The identification of these variants could
lead to insights into how genes predispose individuals to
disease,and might, therefore, inform the development of
improved therapeutic and disease-prevention strategies.
One way to find functionally important variants is
to identify genes that have been acted on by natural
selection.Variants that increase the
FITNESS of an indi-
vidual in its environment might increase in frequency
as a result of positiveselection
(FIG.1),whereas moder-
ately to severely deleterious gene variants tend to be
eliminated by negative (or purifying) selection — a
force to which all genes are probably subject, to main-
tain function.Looking for evidence of positive selection
is an attractive strategy for finding functional variants
because the dispersal of early humans from Africa to
Europe,Asia and the Americas — each with different
climates,pathogens and sources of food — varied the
selective pressures that challenged human populations.
The effects of selection could have been accentuated by
the marked changes in population size,population
density and cultural conditions that accompanied the
introduction of agriculture at the beginning of the
Neolithic period ~10,000 years ago
1,2
.Today, the func-
tional consequences of the genetic variants that facili-
tated survival in ancestral human populations might
underlie the phenotypic differences between individu-
als and groups.So,the analysis of genetic variation in
populations has become central to understanding the
function of genes.
In this review, we highlight some of the types of nat-
ural selection and their effects on the patterns of DNA
variation in the human genome.We explain the relative
strengths and weaknesses of the strategies that can be used
to detect the signatures of natural selection at individual
loci.These strategies are illustrated by their application to
empirical data from the gene variants that are associated
with differences in disease susceptibility.We also outline
the methods proposed to scan the genome for evidence of
selection.Finally,we discuss the problems that are associ-
ated with identifying signatures of selection and with
making inferences about the nature of the selective
process.There are several philosophical issues (for exam-
ple,defining the units of selection),theoretical concepts
and empirical studies in other species, which are beyond
the scope of this review. For further information,the inter-
ested reader is referred to several excellent resources
3–6
.
SIGNATURES OF NATURAL
SELECTION IN THE HUMAN
GENOME
Michael Bamshad*
‡
and Stephen P.Wooding*
During their dispersal from Africa, our ancestors were exposed to new environments and
diseases. Those who were better adapted to local conditions passed on their genes, including
those conferring these benefits, with greater frequency. This process of natural selection left
signatures in our genome that can be used to identify genes that might underlie variation in
disease resistance or drug metabolism. These signatures are, however, confounded by
population history and by variation in local recombination rates. Although this complexity makes
finding adaptive polymorphisms a challenge, recent discoveries are instructing us how and where
to look for the signatures of selection.
FITNESS
The ability of an individual to
reproduce his or her genetic
makeup, which is not always
equivalent to individual
reproductive success.
NATURE REVIEWS | GENETICS VOLUME 4 | FEBRUARY 2003 | 99
*Department of Human
Genetics, University of Utah,
Salt Lake City,
Utah 84112, USA.
‡
Department of Pediatrics,
University of Utah,
Salt Lake City,
Utah 84112, USA.
e-mails:
mike@genetics.utah.edu;
swooding@genetics.utah.edu
doi:10.1038/nrg999
REVIEWS

© 2003
Nature
Publishing
Group
100 | FEBRUARY 2003 | VOLUME 4 www.nature.com/reviews/genetics
REVIEWS
FIXATION
The increase in the frequency of
a genetic variant in a population
to 100%.
BALANCING SELECTION
A selection regime that results in
the maintenance of two or more
alleles at a single locus in a
population.
BACKGROUND SELECTION
The elimination of neutral
polymorphisms as a result of the
negative selection of deleterious
mutations at linked sites.
POLYMORPHISM
The contemporary definition
refers to any site in the DNA
sequence that is present in the
population in more than one
state.By contrast,the traditional
definition referred to an allele
with a population frequency
>1% and <99%.
GENETIC DRIFT
The random fluctuations of
allele frequencies over time due
to chance alone.
selection,compare DNA or amino-acid variation in
populations or species and/or the degree of diver-
gence between them (
BOX 2; TABLE 1;see REF.12 for a
review). The power of these tests is typically deter-
mined by carrying out simulations under a restricted
range of demographic models and parameters to esti-
mate the critical values that support rejection of the
neutral model
13–15
.To this end,an understanding of
population history is crucial for identifying the genes
that are subject to selection.
The confounding effects of population history
Interest in characterizing the patterns of genetic vari-
ation within and among human populations has
grown over the past few years (see
REFS 16,17 for a
review). However, until recently, studies have been
hampered by the relatively small number of poly-
morphic loci that were typed in each individual and
by the restricted sampling of human populations. In
the past few years,many publications have reported
results on the basis of extensive surveys of variation
of the mitochondrial genome
18
, the Y chromosome
19
and various autosomal regions, using microsatellite
or single nucleotide polymorphism (SNP) markers
20
.
These data have allowed broad inferences to be made
Genetic variation: the raw material of selection
Natural selection can act in a population only if muta-
tion has generated heritable genetic differences (or
POLYMORPHISMS) among individuals in the population.
Otherwise,the differences in fitness between individu-
als could not be transmitted from one generation to
the next.In humans,it has been estimated that ~4 new
amino-acid-altering mutations arise per diploid
genome per generation
7
; these mutations can be
broadly categorized as advantageous (that is,adap-
tive),deleterious or neutral (that is,exerting no effects
on the fitness of an individual). The fixation of adap-
tive mutations in a population is an important theme
underlying Darwin’s theory on the origin of species by
natural selection. For the past few decades,however,it
has been widely believed that most genetic variation
— both polymorphisms in species and divergence
between species — is neutral and that polymorphisms
are eliminated or fixed in populations as a conse-
quence of the stochastic effects of
GENETIC DRIFT.The
logic underlying this supposition was outlined by
Motoo Kimura and became known as the neutral the-
ory of molecular evolution
8–11
(BOX 1).
The statistics that are used to summarize poly-
morphism data, and to test for the effects of natural
a Neutral b Positive c Balancing d Background
Time
Figure 1 | Effects of natural selection on gene genealogies and allele frequencies. Each panel (a–d) represents the
complete genealogy for a population of 12 haploid individuals. Each line traces the ancestry of a lineage, and coloured lines trace
all descendants who have inherited an allele that is either neutral (a) or affected by natural selection (b–d) back to their common
ancestor (that is, the coalescence of the genealogy). a | The genealogy of a neutral allele (red) as it drifts to
FIXATION. b | The
genealogy of an allele (green) that is driven to fixation more quickly after the onset of positive selection (arrow) compared with
expectations under a neutral model. Note that the genealogy has a more recent coalescence. c | The genealogy of two alleles
(blue and gold) under
BALANCING SELECTION, which are driven neither to fixation nor to extinction. As a result, the genealogy of the
two alleles has an older coalescence. d | The genealogy of an allele (purple) that drifts to fixation under the influence of
BACKGROUND SELECTION. Each circle represents the elimination of a deleterious (that is, lethal) mutation by background selection.
The coalescence of the lineage is more recent than expected under a neutral model because a linked deleterious mutation caused
the extinction of one lineage (arrow) more quickly than would be predicted.

© 2003
Nature
Publishing
Group
NATURE REVIEWS | GENETICS VOLUME 4 | FEBRUARY 2003 | 101
REVIEWS
pattern estimated from a set of neutral markers that
have been typed in the same individuals or
populations
29,30
.Empirical distributions of the sum-
mary statistics that are sensitive to selection and pop-
ulation history (for example, Tajima’s D,see
BOX 2),
estimated from hundreds of coding and non-coding
regions, are also becoming available
31
.These distribu-
tions can be used to compare a candidate locus to
other regions of the genome to determine whether it
has a pattern of variation that is significantly different.
However,the comparison of results across studies is
often difficult because the distributions are sensitive
to the populations studied and the sampling strategies
used,and these often vary
32
.
The impact of positive selection
In contrast to demographic processes, which affect the
entire genome,natural selection affects specific function-
ally important sites in the genome. In addition, depend-
ing on the local rate of recombination, whenever
selection acts on a mutation it will affect linked sites as
well,leaving its signature in the adjacent chromosomal
region. This signature is manifest as a reduction in varia-
tion at linked sites for two types ofselection. The first —
background selection — removes deleterious mutations
and eliminates variation at linked sites
33
.The strength of
this effect will vary with the recombination rate,the mag-
nitude ofselection and the mutation rate
34
.The second
— genetic hitchhiking — predicts that if a mutation
increases in frequency in a population as a result of posi-
tive selection,linked neutral variation will be dragged
along with it
35
.As a consequence, variation that is not
about the population history of humans, such as the
degree of population subdivision and changes in
population size
21
.
Most of these studies indicate that the human
population has not maintained a constant size,hav-
ing increased from tens of thousands of individuals
to more than 6 billion during the past 100,000
years
22,23
.Furthermore,the human population also
shows substantial
POPULATION STRUCTURE
24,25
.These find-
ings are relevant because population growth and sub-
division can both cause departures from the neutral
model that are indistinguishable from those caused
by natural selection. For example, population genet-
ics theory predicts that the proportion of variants
with a low frequency in a population will increase in
an expanding population,because in such a popula-
tion new mutations are lost at a lower rate
26
.The
excess of low-frequency variants seen in humans for
many types of genetic marker that are presumed to
be neutral has been interpreted as evidence of the
rapid expansion in human population size
23,27
.
However,positive selection can produce a similar
excess of low-frequency variants
28
.
It is also possible for a departure from the neutral
model at any specific locus to be caused by a combi-
nation of both population history and selection. An
important distinction is that demographic processes
similarly affect all loci,whereas the effects of selection
are restricted to specific loci.Therefore,one way that
the confounding effects of population history can be
treated empirically is by comparing the pattern of
variation at a candidate locus with the genome-wide
GENETIC LOAD
The proportion of a population’s
maximum fitness that is lost as a
result of selection against the
deleterious genotypes it
contains.
EFFECTIVE POPULATION SIZE
The size of the ideal population
in which the effects of random
drift would be the same as those
seen in the actual population.
POPULATION STRUCTURE
A departure from random
mating as a consequence of
factors such as inbreeding,
overlapping generations,finite
population size and
geographical subdivision.
Box 1 | The neutral theory of molecular evolution
Before the late 1960s,many evolutionary biologists assumed that most of the polymorphisms in a population were
maintained by balancing selection. However, because the maintenance of balanced polymorphisms was predicted to
impose a large
GENETIC LOAD, most genes were thought to be monomorphic. However,perspectives began to change as the
proliferation of protein sequencing and electrophoresis led to the discovery of extensive amino-acid polymorphisms
both within, and between,species. In a series of papers published during the 1960s and 1970s, Motoo Kimura and others
suggested that the patterns of protein polymorphism seen in nature were more compatible with the hypothesis that
most polymorphisms and fixed differences between species are selectively neutral
8,9
. This proposition was called the
neutral theory of molecular evolution, or the neutral theory.
The development of the neutral theory was motivated by two principal observations.First, between-species
comparisons of amino-acid substitution rates showed that they were regular, or clock-like:although clock-like rates
would be expected if amino-acid substitutions occurred stochastically, they would not be expected if natural selection
were pervasive. Under the pressure of natural selection, amino-acid substitutions would be expected to occur irregularly,
reflecting the unpredictability of environmental change.Second,when the amino-acid substitution rates inferred from
comparisons between species were taken into account,the levels of diversity within species were found to be roughly
proportional to the
EFFECTIVE POPULATION SIZE. Such a pattern would not be expected if natural selection were acting to
balance new variants. Taken together, these patterns were interpreted as evidence that genetic drift, rather than natural
selection, was responsible for maintaining most polymorphisms.
Kimura emphasized that the neutral theory is not incompatible with the idea of an important role for natural selection
in shaping human genetic variation. Strong negative selection, for example,which removes variants from a population,
could affect most new mutations but still have little effect on the levels of polymorphism seen. Some positive selection,
which would sweep mutations to fixation,could also exist without jeopardizing the conclusion that most of the fixed
differences between genes are neutral.The remaining polymorphisms,Kimura argued, represent a mixture of selectively
neutral alleles and mildly deleterious alleles that have not yet been removed by natural selection.
The emphasis of the neutral theory on the importance of genetic drift changed the focus of population genetic
analysis. It transformed commonly held views on the role of natural selection in maintaining variation, introducing a
more sophisticated outlook on the balance of selection and drift, which persists in present evolutionary theory.

© 2003
Nature
Publishing
Group
102 | FEBRUARY 2003 | VOLUME 4 www.nature.com/reviews/genetics
REVIEWS
coding region of a gene is compared between humans
and one or more other species to determine whether the
rate of amino-acid substitution is higher or lower than
expected (
BOX 2;see REFS 42,43 for review). The current
wealth of DNA sequence data is making this approach
increasingly popular, and an emphasis has been placed
on comparing loci between humans and chimpanzees
for evidence of positive selection
44,45
.The identification
of such loci might provide insights into the nature of
changes,such as the origin of speech,which led to the
evolution of modern humans
46
(see REF.47 for a review).
In contrast to the investigation of selection among
species,relatively few studies of local positive selection
in humans have been carried out, even though the loci
that show differential selection across populations
might be determinants of disease resistance.Such stud-
ies might be useful if common diseases,such as diabetes,
obesity and atherosclerotic heart disease, are caused by
genetic variants that were positively selected in ancestral
environments but are detrimental at present (for exam-
ple,the so-called ‘thrifty’genotypes)
48
.Studies of local
positive selection can be designed to screen the entire
genome for regions that are affected by selection (see
below),or they can focus on testing specific candidate
genes.The testing of specific candidates is limited by our
lack of knowledge about the most suitable genes for
linked to the adaptive mutation is eliminated,resulting in
a
SELECTIVE SWEEP (FIG.2).Therefore,models predict that
genetic hitchhiking will cause a greater overall reduction
in genetic diversity, and that the effect will be more pro-
nounced in regions of lower recombination. Both types
ofselection will result in an overall positive correlation
between genetic diversity and recombination rate if the
strength and frequency of positive and/or background
selection are sufficiently high throughout the genome
36,37
.
Empirical data from several plant and animal species,
including mice
38
and fruitflies
37
— the best-studied
species so far — are consistent with the predictions of
recurrent selective sweeps.In humans,it seems that about
half of the
VARIANCE in nucleotide diversity might be
explained by the local recombination rate
39
,although
more comprehensive studies are needed.These results
indicate that selection might have been more important
in shaping patterns of variation in the genome than was
previously anticipated,although the relative importance
ofbackground selection and genetic hitchhiking remains
unknown.Indeed,it is possible that both processes con-
tribute to the patterns of variation
40,41
.
Most of what is known about the impact of adaptive
evolution on the human genome comes from studies of
the patterns of genetic differences between humans and
other species.In general, the genetic variation in the
SELECTIVE SWEEP
The process by which positive
selection for a mutation
eliminates neutral variation at
linked sites.
STANDARD NEUTRAL MODEL
A hypothetical panmictic
(randomly mating) population
of constant size in which genetic
variation is neutral and follows a
model (the ‘infinite sites model’)
in which each new mutation
occurs at a site that has not
previously mutated.
VARIANCE
A statistic that quantifies the
dispersion of data about the
mean.
Box 2 | Measures of genetic variation
Several descriptive statistics are commonly used to summarize polymorphism data in a sample of DNA sequences.For
example,
θ
W
describes the proportion of segregating sites in a sample (corrected for the size of the sample)
112
,and
nucleotide sequence diversity (
π
) describes the mean number of differences per site between two sequences chosen at
random from a sample of sequences
113
. The average
π
in humans (~7.5 ×10
−4
) is relatively low
114,115
,although
π
can vary
by more than an order of magnitude among genomic regions
116
. Each of these statistics is an estimator of the population
mutation rate,
θ
= 4N
e
µ
,where N
e
is the effective population size and
µ
is the neutral mutation rate per generation.
Therefore,
π
can also be used to estimate the N
e
of humans,which, on the basis of diverse genetic data, is ~10,000
(REFS 27,117). This is smaller than our census size and indicates that probably only those polymorphisms that had
substantial effects on fitness were likely to have overcome the effects of genetic drift.
Departure from a
STANDARD NEUTRAL MODEL can be assessed using several test statistics that use comparisons of
estimators of
θ
(TABLE 1). One frequently used test statistic,Tajima’s D,compares the estimates of
θ
W
and
π
from a single,
non-recombining region of the genome
118
. The difference between
θ
W
and
π
is expected to be zero under the neutral
model, and so a non-zero Dis a sign of a departure from the neutral model due to a relative excess or deficiency of
polymorphisms of various frequencies.For example,background selection will eliminate polymorphisms linked to
deleterious alleles, allowing them to reach only low frequencies,whereas positive selection will tend to eliminate older,
high-frequency alleles, and newer,low-frequency alleles will hitchhike with the target of selection
(FIG.1).As a
consequence,positive or background selection can to lead to an excess of polymorphisms at low frequencies.
Other estimators of
θ
vary in their sensitivity to polymorphisms of different frequencies.For example,alleles that have
been targets of recent positive selection might exist at a relatively high frequency. A recently developed estimator,
θ
H
,is
more sensitive to such alleles, and therefore the test statistic based on it, H, might be more powerful for detecting recent
positive selection
119,120
. It is worth noting that all of these test statistics use the allelic distribution and/or the level of
allele variability, both of which are dependent on the genealogy of a locus. Therefore,they rely on assumptions about the
demographic histories of the populations in which the samples were ascertained.As a consequence, the interpretation of
the results of these test statistics can be challenging, and they rarely provide unambiguous evidence of selection.
By contrast,tests for selection that are independent of the genealogy of a locus have provided clear evidence for
selection. These tests use comparisons of the variability and divergence of different types of mutation at a locus, such as
the rate of non-synonymous mutations (d
N
) versus the rate of synonymous mutations (d
S
) (TABLE 1). In this example, a
significantly increased rate ratio (d
N
/d
S
) has indicated that human olfactory receptor genes
121
, human leukocyte antigen
(HLA) loci
101
and breast cancer 1,early onset (BRCA1)
122
have all been subject to positive selection. These tests are
generally conservative,because the substitution rates are averaged across all the amino-acid sites tested.Alternative
strategies include testing functionally important domains of a protein individually, testing single amino-acid sites
123
or separately testing the lineages in a phylogenetic tree.

© 2003
Nature
Publishing
Group
NATURE REVIEWS | GENETICS VOLUME 4 | FEBRUARY 2003 | 103
REVIEWS
gorillas — the two species with which humans most
recently shared a common ancestor — are commonly
used as outgroups when examining human frequency
spectra. Under a neutral model,this distribution has a
characteristic shape,which can be skewed by natural
selection
(FIG.2).If we know which allele is ancestral and
which allele is derived at each site,we can make infer-
ences about the type of selection that has affected a locus.
Although positive selection is expected to skew the spec-
trum towards an excess of low-frequency alleles, such a
skew is not generally expected in regions that are subject
to background selection
28
.
In the case of four of the SNPs in CYP1A2,the minor
allele in humans was the fixed allele in chimpanzees and
gorillas, so the common SNPs in humans were inferred to
be the derived state (even though they each had a fre-
quency >90%).By comparison with the expected frac-
tion ofvariants at each frequency estimated under the
neutral model, the site frequency spectrum for CYP1A2
showed an excess of both low- and high-frequency alleles.
This indicates that CYP1A2might have been influenced
by both positive selection and recent population growth,
although the relative strengths of each are unclear.
Gene genealogies and selection
Natural selection can also affect the genealogy of alleles,
the relationships of which can be depicted in a tree or
network
(FIG.2a). The parameters of the genealogical
process that have given rise to a tree can be estimated
using coalescence theory
(BOX 3). Departures from the
neutral model are thought to reflect the effect that pop-
ulation history and natural selection have had on the
shape of the genealogy (see
REF.70 for a review).For
example, positive selection that sweeps an adaptive
variant to fixation can distort the genealogy to create a
star-like pattern
71
,which is a sign of an excess of low-
frequency variants that are connected to a common
ancestor by branches with similar, often short, lengths.
Coalescence theory can be used to test whether natural
selection has produced such a genealogy.
analysis.However, our increasing knowledge of the bio-
logical mechanisms that underlie phenotypic traits has
led to the accumulation of circumstantial evidence,
indicating that certain loci might have been the targets
of selection.This has led to notable recent successes in
finding signatures of local positive selection in human
populations (for examples,see
REFS 49–66).In turn,a
review of the properties of these loci might refine our
strategies for finding the signatures of selection.
Selection and the site frequency spectrum
Among the most promising candidates to test for a sig-
nature of local positive selection are genes that encode
the proteins that are involved in drug transport and
metabolism.Many of these genes show marked differ-
ences in allele frequencies between populations,and
gene variants have been associated with variable
responses to foods and drugs
67
. An example is
cytochrome P450 1A2 (CYP1A2),which encodes an
enzyme that oxidizes carcinogenic arylamines,aceta-
minophen and several widely prescribed anti-psychotic
drugs
68
. The hepatic mRNA expression of CYP1A2
varies by as much as 15-fold among individuals and,
accordingly, variants of CYP1A2 might underlie inter-
individual variation in cancer susceptibility, as well as
variation in response to toxins or medications
69
.The
regulatory region of this gene was a logical candidate to
screen for a signature of selection.
Analysis of 3.7 kb of a non-coding DNA sequence 5′
of CYP1A2 in Africans, East Asians and Europeans
showed a pattern of SNP frequencies that indicated
recent positive selection
49
.To illustrate this, we focus on
the site frequency spectrum.In most spectra, each
MINOR
ALLELE
is inferred to be the derived (that is,the younger or
more recent) allele,because derived alleles typically exist
at lower frequencies than ancestral ones. Moreover,if
positive selection was recent,linked derived variants
might not be fixed,and so exist at higher frequencies
than expected.This can be inferred by comparing the
human sequence to an
OUTGROUP. Chimpanzees and
MINOR ALLELE
The less frequent of two alleles at
a locus.
OUTGROUP
A closely related species that is
used for comparison, for
example, to infer the ancestral
versus the derived state of a
polymorphism.
Table 1 | Commonly used tests of neutrality
Test Compares References
Tests based on allelic distribution and/or level of variability
Tajima’s D The number of nucleotide polymorphisms with 118
the mean pairwise difference between sequences
Fu and Li’s D, D* The number of derived nucleotide variants observed only once 129
in a sample with the total number of derived nucleotide variants
Fu and Li’s F, F* The number of derived nucleotide variants observed only once in 129
a sample with the mean pairwise difference between sequences
Fay and Wu’s H The number of derived nucleotide variants at low and high 119
frequencies with the number of variants at intermediate frequencies
Tests based on comparisons of divergence and/or variability between different classes of mutation
d
N
/d
S
, K
a
/K
s
The ratios of non-synonymous and synonmyous 130,131
nucleotide substitutions in protein coding regions
HKA The degree of polymorphism within and 132
between species at two or more loci
MK The ratios of synonymous and non-synonymous 128
nucleotide substitutions in and between species
HKA, Hudson–Kreitman–Aguade; MK, McDonald–Kreitman.

© 2003
Nature
Publishing
Group
104 | FEBRUARY 2003 | VOLUME 4 www.nature.com/reviews/genetics
REVIEWS
generate hypotheses about the history of a locus, it must
be noted that they should not be used,alone, to make
inferences of selection.
Linkage disequilibrium and selection
Most of the strategies that are used to detect whether
natural selection has affected a specific allele measure
the departure of the frequency of the allele from
expectations under a neutral model, with the assump-
tion that there has been no local recombination.
Indeed,recombination has often been considered a
nuisance in this context.It is,therefore,ironic that the
block-like nature of
LINKAGE DISEQUILIBRIUM
(LD) across
the human genome offers a new way to detect a signa-
ture of recent positive selection
50,73
.The logic underly-
ing this strategy is straightforward.When a mutation
arises, it does so on an existing background haplotype
characterized by complete LD between the new muta-
tion and the linked polymorphisms
(FIG.4). Over time,
new mutations and recombination reduce the size of
this haplotype block such that,on average,older and
relatively common mutations will be found on
smaller haplotype blocks (that is, there is only short-
range LD between the mutation and linked polymor-
phisms).Younger,low-frequency mutations might be
associated with either small or large haplotype blocks.
A signature of positive selection is indicated by an
allele with unusually long-range LD and high popula-
tion frequency.The formal implementation of this
strategy has recently been introduced as the long-
range haplotype (LRH) test
50
.
Glucose-6-phosphate dehydrogenase (G6PD) is one
of several examples of genes in which alleles that are
common in sub-Saharan Africans have been associated
with resistance to infection
74–76
. G6PD is the only
enzyme in red blood cells that can recycle nicotinamide
adenine dinucleotide phosphate (NADP), which is
needed to prevent oxidative damage to the cell.
Hundreds of G6PD variants have been identified,
although most of them are relatively uncommon
77
.
G6PD-202A reduces enzyme activity to ~10% of its
normal value, and is found at frequencies as high as
25% in sub-Saharan African populations
78
.This variant
is advantageous in certain environments, because it
reduces the risk of malarial disease by 40–60% in het-
erozygous females and hemizygous males
79
.
Haplotypes that bear G6PD-202A have significantly
less microsatellite variability than predicted by a coales-
cence model.This low level of variability, in conjunction
with the high frequency of G6PD-202A, indicates that
G6PD-202Amight have risen in frequency so rapidly
that there was no time to accumulate new variation in
nearby polymorphisms
51,52
.Long-range LD around hap-
lotypes that bear G6PD-202Aextends for hundreds of
kilobases
(REF.50),which is significantly longer than the
LD of other G6PDvariants of comparable frequency.
Both of these patterns differ from the pattern of haplo-
type variation and LD seen at other loci in the same pop-
ulations,providing molecular evidence that G6PD-202A
has been a target of recent positive selection. The date of
origin of G6PD-202A estimated from these data ranges
The use of LIKELIHOOD ANALYSIS,based on the coales-
cent, to test hypotheses of selection in humans is
becoming more common
54,55
.In the meantime,most
human polymorphism data are still analysed using
summary statistics based on coalescence theory, as well
as methods adopted from
PHYLOGENETICS.In the latter
case, once polymorphism data are available, the
genealogical tree of the locus is estimated and compared
with inferences made using other methods.This strat-
egy has practical value,particularly if questions about
the haplotype structure of a locus are being explored. In
a tree of CYP1A2 haplotypes, for example, the common
haplotype that shares the four high-frequency-derived
variants (asterisk in
FIG.3a) is connected to other haplo-
types by short branches, in a star-like pattern
(FIG.3a).
This pattern is reminiscent of a tree distorted by positive
selection
(FIG.2).This tree can also be used to facilitate
the cloning of functional variants,because haplotypes
sharing functional variants that are associated with the
same phenotypic trait generally share a common inter-
nal branch in the network.The sequences of these hap-
lotypes can be compared with one another to identify all
of the mutations shared exclusively among them.Each
of these mutations can then be tested,alone or in com-
bination,to determine whether they are of functional
significance
72
,thereby reducing the number of muta-
tions that need to be screened for functional effects.
Although the topology of these networks can help
LIKELIHOOD ANALYSIS
A statistical method that
calculates the probability of the
observed data under varying
hypotheses, in order to estimate
model parameters that best
explain the observed data and
determine the relative strengths
of alternative hypotheses.
PHYLOGENETICS
Reconstruction of the
evolutionary relationships (that
is, the phylogeny) of a group of
taxa, such as species.
LINKAGE DISEQUILIBRIUM
(LD). The non-random
association of alleles in
haplotypes.
Box 3 | The coalescent process
If a sample of genes is
collected,and the ancestry
of those genes is traced
backwards in time, common
ancestors will be
encountered.As these
common ancestors are
encountered,the number of
distinct ancestral lineages
decreases from n(the original
sample size) to n−1,n −2
and so on, eventually ending
in one lineage, which is the
common ancestor of the
entire sample; this is called
the coalescent process.The
diagram shows the genealogy
of a population of 12 haploid
individuals. The red lines
trace the ancestry of their 12
lineages back to the most
recent common ancestor.
The coalescent process is a
point of key interest in efforts to detect natural selection in human genes. Because natural
selection can affect the rate at which common ancestors are encountered (for example,
compare the time at which the common ancestor is encountered in each panel of
FIG.1),
the shape of gene genealogies can inform us about the selective processes that have been
involved.Selective sweeps,for example,result in ‘shallow’gene genealogies with a few, rare
mutations.By contrast,balancing natural selection results in ‘deep’ gene genealogies in
which many mutations are found at intermediate frequencies.
Number of
ancestral lineages
1
2
3
4
5
12

© 2003
Nature
Publishing
Group
NATURE REVIEWS | GENETICS VOLUME 4 | FEBRUARY 2003 | 105
REVIEWS
A mutation that produces a Cys–Tyr substitution at
position 282 of the mature polypeptide (C282Y)
accounts for ~85% of all haemochromatosis muta-
tions
82
.This mutation is almost exclusive to individuals
of European descent
83
,in whom it has a frequency of
5–10%.An assessment of LD between the C282Y muta-
tion and linked polymorphisms showed a substantial
degree of LD, indicating that the mutation probably
arose only ~60 generations ago
84,85
.It is intriguing that
C282Y has reached a relatively high frequency in such a
short period of time. A coalescence-based analysis of the
frequency of C282Y and its age (based on the level of
the LD between the mutation and the linked polymor-
phisms) showed that it is improbable that the observed
frequency of C282Y could have been achieved by
genetic drift alone
53
.Instead,recent positive selection is
probably responsible for the high frequency of the
mutation in Europeans.
Support for the hypothesis of a selective sweep
by C282Y comes from data on iron deficiency among
from ~2,500 to 6,500 years ago
50,51
.Interestingly,these
dates are in agreement with archaeological data that
indicate that malaria might have had a substantial
impact on sub-Saharan Africans only in the past 10,000
years,concordant with a recent expansion in an already
large effective population of Plasmodiumfalciparum
80
and with the diversification of Anopheles gambiae,a
mosquito vector of malaria
81
.
Evidence of local positive selection has also been
found outside Africa.Idiopathic haemochromatosisis
an autosomal-recessive disorder caused by mutations in
the HFE gene, which is characterized by excessive
intestinal iron uptake.Iron accumulates in the heart,
liver, pancreas,joints and skin; this can lead to hepatic
cirrhosis, diabetes, arthropathy and heart failure.
Because iron accumulates slowly in affected individuals,
disease symptoms are not usually seen until the fifth or
sixth decade in males,if at all.Among females, the age of
onset is even later because of iron loss due to menstrua-
tion,pregnancy and lactation.
HAPLOTYPE
The combination of alleles or
genetic markers found on a
single chromosome of a given
individual.
SITE FREQUENCY SPECTRUM
The fraction of polymorphic
sites at which a minor or derived
allele is present in one copy, two
copies and so on.
123456
2
4
6
8
10
12
14
= 0.076
D = 0.06
π
Sites
Occurences
123456
2
4
6
8
10
12
14
= 0.085
D = 0.40
Sites
Occurences
123456
2
4
6
8
10
12
14
= 0.063
D = –0.89
Sites
Occurences
a Genealogies b Haplotypes c Site frequency spectra
Locus under
positive selection
Locus under balancing selection
Locus under
no selection (neutral)
π
π
Figure 2 | The effects of selection on the distribution of genetic variation. a | The genealogies of three genes that are typical of
loci under positive selection (top), balancing selection (middle) and no selection (neutral) (bottom) are depicted. Each circle
represents a mutation, and the colour shows the final frequency of each mutation in the sampled
HAPLOTYPES (b) and the SITE
FREQUENCY SPECTRA
(c). For each gene, the number of segregating sites is 20. b | Each haplotype contains mutations that have
accumulated on each lineage in the gene genealogy, assuming no recombination. c| The site frequency spectrum of each gene.
Positive selection (top) can result in a lower level of sequence diversity (
π
), an excess of low-frequency variants (red) and,
consequently, a negative value of Tajima’s D
(BOX 2). Balancing selection (middle) can result in a higher level of sequence diversity
(
π
), an excess of intermediate-frequency variants (purple) and, consequently, a positive value of Tajima’s D. The diversity estimate
and site frequency spectrum of a neutral locus (bottom) can be used for comparison.

© 2003
Nature
Publishing
Group
106 | FEBRUARY 2003 | VOLUME 4 www.nature.com/reviews/genetics
REVIEWS
β-globin
54
and the DUFFY BLOOD GROUP
loci
56,57
.At least one
gene that contributes to the diversification of morpho-
logical traits among humans — the melanocortin-1
receptor (MC1R) — seems to have been under various
selective pressures
58,59
.Evidence for positive selection has
also been found in genes that encode the dopamine
receptor D4 (DRD4)
60
,calpain-10
(REF.29), factor IX
(REF.61),CD40 ligand
50
,dystrophin
62
,monoamine oxi-
dase A (MAOA)
63
,lactase-phlorizin hydrolase
64
and
chemokine (C-C motif) receptor 5 (CCR5)
65,66
.Most of
these genes were candidates for study because of our
knowledge of their biology. Relatively little is known
about most of the 30,000 or so genes in the human
genome.Although it is easy to begin developing ad hoc
stories about how selection might have influenced a
candidate,it is prudent to consider each candidate with
a degree of scepticism. Many differences between the
genes of humans and other species have been affected
by selection,but it is less clear how many genes have
been affected by local adaptive processes,and much less
clear whether these genes are important for understand-
ing human phenotypic variation.
Balancing selection
Natural selection does not always increase or decrease
the frequency of a single allele at a locus.Sometimes,
selection tends to maintain two or more alleles at a locus
individuals who are heterozygous for a haemochro-
matosis mutation.A study of >1,000 American het-
erozygotes showed that 32% of normal homozygous
females of reproductive age had iron deficiency (defined
as a serum ferritin level <12 µg l
−1
),compared with only
21% of haemochromatosis heterozygotes
86
.Among
males over the age of 18 years,the corresponding per-
centages were 4% and 2%.Iron deficiency,which might
have been more common in earlier populations,is asso-
ciated with an increased risk of preterm delivery and
with low birth weight
87
.Therefore,the C282Y allele
might have improved the fitness of both male and
female heterozygous carriers.Because of the late age of
onset of symptoms,an adverse effect in homozygotes
might have had less of an effect on fitness. It is worth
noting, however,that even mutations having an adverse
effect later in life might have weakly deleterious effects
early in life,that could alter the frequency distribution
of an allele
88
. Alternatively, the ‘Grandmother hypothe-
sis’suggests that mutations with late-onset deleterious
effects could have strong effects if they impair the ability
of their host to care for its descendants, indirectly
diminishing the fitness of the host
89
.
Evidence of positive selection acting on other genes
is slowly beginning to accumulate.Signatures of positive
selection have been found in genes that have been used
as classical models of selection, including those at the
SINGLE-LINKAGE JOINING
ALGORITHM
A simple clustering algorithm
that begins with all data points
(for example, haplotypes) in
separate clusters,and then
iteratively joins pairs of similar
clusters.
DUFFY BLOOD GROUP
This group is defined by variants
in a chemokine receptor that is
present on the surface of several
types of cell,including red blood
cells.This receptor must be
present for Plasmodium vivax to
invade cells and cause malaria.
Chimpanzee
(38)
Chimpanzee
(10)
Europe Africa Asia
a
b
*
*
Figure 3 | Using phylogenetic techniques to infer
haplotype structure. The diagram shows networks of
cytochrome P450 1A2 (CYP1A2) (a) and chemokine receptor
(C-C motif) 5 (CCR5) (b) haplotypes. Each network was
generated using a
SINGLE-LINKAGE JOINING ALGORITHM. Each
haplotype is represented by a circle, and the area of the circle is
proportionate to the haplotype frequency. Branches that connect
haplotypes represent one nucleotide difference unless crossed
by hatch marks, which denote the number of mutations on that
branch. Numbers in parentheses indicate the difference between
the chimpanzee haplotype and the most similar human
haplotype. a| In the network of haplotypes of the upstream non-
coding region of CYP1A2, the more derived haplotypes, which
are separated from the chimpanzee by a long branch, are found
at higher frequencies than expected under a neutral model.
This pattern is consistent with the effects of a selective sweep.
b | Balancing selection on the 5′cis-regulatory region of CCR5
results in a haplotype network with two main clusters of high-
frequency haplotypes separated by relatively long branches.
Chimpanzee haplotype is indicated by the circled black dot.

© 2003
Nature
Publishing
Group
NATURE REVIEWS | GENETICS VOLUME 4 | FEBRUARY 2003 | 107
REVIEWS
because alleles are kept in more equal frequencies com-
pared with a neutral locus, and an advantageous allele
that is introduced to a population by migration will be
positively selected.This increases the chances of survival
of an advantageous allele compared with a neutral allele.
Both of these processes are expected to decrease popula-
tion differentiation,commonly measured using
WRIGHT’S
FIXATION INDEX
(F
ST
). The F
ST
estimate for the 5′ cis-
regulatory region of CCR5 is 1.6%,which is nearly ten-
fold lower than the typical F
ST
estimates of 10–15% that
are found for other regions of the genome among vari-
ous populations throughout the world
16
.
Because balancing selection maintains two or more
lineages over a longer period of time than expected,the
genealogy of the locus is expected to differ from that of a
neutral locus
100
.Genealogies are generally characterized
by two or more classes of lineages that are separated by
relatively long branches
(FIGS 1c, 2),a pattern recapitu-
lated in the tree of CCR5haplotypes
(FIG.3b).Similar
genealogies can be caused by population subdivision, in
which haplotypes are restricted to specific populations.
The length of each branch is,however, ultimately
dependent on the age of the mutation and the length of
time that it has been under balancing selection. If a
mutation has arisen relatively recently,or the onset of
balancing selection is recent,both of which might be the
case for G6PD, the branch lengths might be short.
Branch lengths for mutations that are subjected to
recent positive selection, such as CCR5-
∆
32 in
Europeans (see asterisk in
FIG.3b),will also be short. The
latter shows that more than one type of selection can
affect the pattern of genetic variation at a locus.Whereas
local positive selection has recently increased the fre-
quency of CCR5-
∆
32in Europeans,polymorphisms in
the 5′cis-regulatory region of CCR5 that are associated
with disease progression have been maintained by bal-
ancing selection.This pattern is similar to that seen at
the major histocompatibility complex (MHC) locus, in
which allele frequency variation has been affected by
both positive and balancing selection
101
.
Balancing selection involves rare-allele advantage.
Two types of selection that feature rare-allele advantage
are negative frequency-dependent selection and gener-
alized overdominance.In negative frequency-dependent
selection,the fitness of an allele decreases as it becomes
more common.In generalized overdominance,het-
erozygotes maintain a selective advantage over homozy-
gotes,and,therefore,the rare alleles benefit from their
representation in the heterozygotes.This latter type of
selection is thought to maintain the high levels of allelic
variation seen at the MHC locus
102
,an insight derived
from functional data showing that MHC heterozygotes
can present an expanded spectrum of antigens to T cells
compared with that of MHC homozygotes.
Many coding regions in the human genome do
not have an excess of low-frequency alleles.This indi-
cates that balancing selection might be more com-
mon than is generally perceived
55
. Although this
point of view was previously common,it lost support
for several reasons.One objection was that a large
number of loci maintained by overdominance would
in a population (FIG.2).This is known as balancing selec-
tion because the frequencies of alleles are maintained in
a balance,often as a result of a rare allele advantage.
Balancing selection can,therefore,maintain an excess of
alleles at intermediate frequencies, and variation at
linked loci can also accumulate because of genetic hitch-
hiking
90,91
.In many plants and animals,balancing selec-
tion seems to be involved in maintaining diversity at the
loci that coordinate recognition between self and non-
self
92
.In humans,this has been best studied at loci that
are involved with host–pathogen responses,including
human leukocyte antigen (HLA) class I and class II
genes
93
,β-globin
54
,G6PD
94
, glycophorin A
95
,interleukin
4 receptor-α
96
and CCR5(REF. 30).
CCR5 is a seven-transmembrane G-protein-coupled
chemokine receptor that,along with CD4, is required on
the surface of a cell for the entry of the human immun-
odeficiency virus type 1 (HIV-1).The role of CCR5 in the
pathogenesis of AIDS was highlighted by the observation
that a small fraction of individuals that are resistant to
infection by HIV-1 are homozygous for a 32-bp deletion
(CCR5-
∆
32) in its open reading frame (ORF),which
eliminates the cell-surface expression of CCR5
(REF.97).
This allele is found at unusually high frequencies only in
populations from North-eastern Europe, where it seems
to have been a target of local positive selection
65,66
.
Nevertheless, most polymorphisms in CCR5 that are
associated with HIV-1 disease progression are in the 5′
cis-regulatory region that flanks the ORF.As in the HLA
genes, genetic diversity in this region is higher than
expected,with a site frequency spectrum characterized by
an excess of intermediate frequency alleles
30
(FIG.2).
Loci that are subjected to balancing selection,which
favours intermediate-frequency alleles,are expected to
show a different pattern of sequence diversity compared
with neutral loci
98,99
. Balancing selection increases
within-population diversity relative to total diversity,
b
c
Positive selection
Neutral
a
Time
Time
Figure 4 | Detecting recent positive selection using linkage disequilibrium analysis.
a | A new allele (red) exists at a relatively low frequency (indicated by the height of the red bar)
on a background haplotype (blue) that is characterized by long-range linkage disequilibrium (LD)
(yellow) between the allele and the linked markers. b | Over time, the frequency of the allele
increases as a result of genetic drift, and local recombination reduces the range of the LD
between the allele and the linked markers (that is, it creates short-range LD). c| An allele
influenced by recent positive selection might increase in frequency faster than local recombination
can reduce the range of LD between the allele and the linked markers.
WRIGHT’S FIXATION INDEX
(FST). The fraction of the total
genetic variation that is
distributed among
subpopulations in a subdivided
population.

© 2003
Nature
Publishing
Group
108 | FEBRUARY 2003 | VOLUME 4 www.nature.com/reviews/genetics
REVIEWS
survey of >5,000 microsatellites typed in a sample of
Europeans identified more than 40 regions with extreme
skews in the site frequency spectrum
105
.Some of these
regions are likely to have been affected by selection.The
question of whether these regions contain genes that
have been a target of positive selection is,however, com-
plicated by our limited understanding of the impact of
selection on microsatellite variability.Additionally,the
distance over which hitchhiking is detectable might be
too small,depending on the local rate of recombination,
compared with the average physical distance between the
microsatellites tested and the genes that might have been
influenced by selection
109
.This limitation indicates that a
more efficient strategy might be to assay markers in or
near genes.
A recent example of an approach in which markers
in or near genes were analysed,examined ~9,000 SNPs
and identified the outliers in the extreme tails of the
empirical distributions F
ST
(FIG.5),an approach that does
not depend on assumptions about population history.
SNPs with F
ST
patterns that indicated that they had been
subject to natural selection identified 174 candidate
genes
108
,including peroxisome proliferative activated
receptor-γ (PPARG),which has been associated with
type II diabetes.This screen could be considered conser-
vative because the SNP data set that was analysed might
have been over-represented with common SNPs,which
are expected to be shared across populations. Therefore,
these SNPs would be expected to have smaller differ-
ences in allele frequencies between populations.
In general,the results of screens for signatures of
selection have been similar, with a small percentage of
loci seeming to deviate substantially from expectations.
It remains unclear,however, what proportion of these
genes have been targets of positive selection.The next
step will be to test the predictions of these screens,
perhaps by more direct tests of selection, such as the
impose an excessive genetic load on a population
103
.
However,recent models show that the number of loci
that could be maintained by overdominance without
a substantial genetic load far exceeds the number of
loci in the human genome
104
.
Scanning the genome for natural selection
The approaches that are used to detect selection at indi-
vidual candidate loci are, in principle, adaptable to
scanning the entire human genome for signatures of
selection — albeit with the same limitations imposed by
the effects of population history
(BOX 4).In addition,
these scans might improve our understanding of the
impact of natural selection across the entire genome.
Such scans would rely on information from genetic
markers that were assayed in a representative set of indi-
viduals from selected populations.Although neutral
regions would be expected to have a similar pattern of
variability and allele frequency distribution among pop-
ulations,the patterns of variation in regions that are
affected by selection might differ
(FIG.5). For example, an
excess of rare alleles in a region, or more than expected
differentiation among populations at a marker (that is,a
high F
ST
),might be a signal of recent positive selection.A
region that is characterized by an excess of intermediate
frequency alleles,or by less than the expected differentia-
tion among populations at a marker (that is,a low F
ST
),
might be under balancing selection.
The wealth of nucleotide polymorphism data that
has become available during the past few years has pro-
vided an exciting opportunity to carry out genome scans
for selection.Several scans of the human genome have
been undertaken to search for regions under natural
selection,and more are underway
31,105–108
.These scans
vary in strategy by, for example,using either a battery of
microsatellites versus SNPs, or screening anonymous
regions of the genome versus coding regions.A recent
2 ×2 CONTINGENCY TABLE
A 2 ×2 table that describes the
cross-classification of data that
are divided into two groups with
two categories in each.
Box 4 | Approaches to scanning the genome for selection
Many analytical approaches have been developed to screen a genome for evidence of selection
124
. In general, for each of
several loci, these tests compare the degree of differentiation among sample populations with the overall level of diversity.
Under a neutral model,the variation in allele frequencies among populations is determined by genetic drift alone, the
impact of which depends only on the demographic history of a population. Therefore,all loci are expected to show the
same degree of differentiation.If positive selection has increased the frequency of an allele in one population, but not the
other, a higher fraction of variation will be distributed between populations. Positive selection will also tend to reduce the
total variability in a population. Both effects of positive selection are expected to increase the level of differentiation and,
consequently,the F
ST
between populations.
The first approach that was developed to screen multiple loci was the Lewontin–Krakauer test,which used the variance of
F
ST
values among loci to identify those with an F
ST
that deviated more than was expected
125
.This test was criticized for being
unreliable because,given certain population histories, it inflated the expected variance.This led to the development of more
refined tests that are robust over a wide range of demographic models
126,127
. These tests use comparisons of the relationship
of F
ST
to heterozygosity,estimates of genetic distance between populations or reduced variability at a locus compared with
expectations under a neutral model or various demographic models
(FIG.5). For each of these tests,the ability to detect
signatures of selection depends on the marker density, the distance between markers and the site under selection,the local
recombination rate,the strength of selection and the assumptions made about population history.
An alternative strategy to screen the genome for selection is to use tests that do not depend on assumptions about
population history. One example is the McDonald–Kreitman test,in which the d
N
/d
S
of polymorphisms in species is
compared with the d
N
/d
S
of fixed differences between species in a 2 ×2 CONTINGENCY TABLE
128
. If polymorphism and
divergence are the consequence of only mutation and drift, the ratio of the number of fixed differences to polymorphisms is
the same for both non-synonymous and synonymous mutations.

© 2003
Nature
Publishing
Group
NATURE REVIEWS | GENETICS VOLUME 4 | FEBRUARY 2003 | 109
REVIEWS
subdivision can be mixed up with selection.Moreover,
the expectations of population genetic models are depen-
dent on assumptions about demographic parameters for
which estimates remain ambiguous. Therefore,contin-
ued progress towards identifying the genes that are sub-
ject to selection will depend on understanding more
about the demographic structure of human populations.
Our interpretation of how selection has influenced
the human genome is relatively simple,whereas, in
fact,the effects of selection will probably be complex.
Selection intensity can fluctuate over time, and genetic
drift might dominate in some populations, whereas
selection might dominate in others. Most population
genetic models of positive selection assume that a neu-
tral site is linked to only one functional site on which
selection is acting.Whether this is correct depends on
the rate at which positive selection sweeps mutations
to fixation.If this is frequent,predictions made on
the basis of these models might not be accurate.
comparison of non-synonymous to synonymous sub-
stitutions.Some of the genes that deviate from expecta-
tions are members of the same gene family. In some
cases,these within-family trends might be due to the
clustering of genes with similar function to the same
region of the genome,or to the co-evolution of genes
that interact with one another.Genes that encode medi-
ators in the same metabolic pathway or developmental
programme might show a similar signature of selection.
Even after a candidate locus has been shown to be sub-
ject to selection,a substantial amount of work will be
required to identify the causal variants and to under-
stand their relationship to a human phenotype.
Conclusions
Our inferences of signatures of selection are constrained
by an insufficient understanding of population demogra-
phy and of local rates of recombination. Patterns of
nucleotide variability caused by population growth and
0 0.7
0
0.25
R
ST
Fraction of loci
a 377 STR loci
0 0.7
0
0.25
F
ST
Fraction of loci
b 100 Alu loci
0 0.7
0
0.9
R
ST
Heterozygosity
0 0.7
0
0.9
F
ST
Heterozygosity
Figure 5 | Screening the human genome for signatures of natural selection. Heterozygosity and F
ST
or R
ST
values (R
ST
is an
analogue of F
ST
designed for microsatellites) for two large data sets of markers that are distributed throughout the human genome.
a | A set of 377 microsatellites (short tandem repeats) typed in 958 individuals from Africa, Asia and Europe
24
. b | A set of 100 Alu
insertion polymorphisms typed in 207 individuals from Africa, Asia and Europe
25
. Left-hand panels: one strategy to find loci that have
been subject to natural selection is to identify outliers in the empirical frequency distribution of F
ST
or R
ST
. A high F
ST
indicates that the
region might have been subject to local positive selection, whereas a low F
ST
can be seen in regions under balancing selection (see
text for details). Right-hand panels: another strategy to identify regions that are affected by selection is to identify outliers in a plot of
F
ST
or R
ST
versus heterozygosity. Loci in regions that are far from the origin in each plot, including markers with exceptionally high
heterozygosity, exceptionally high F
ST
(or R
ST
) or both, are the most obvious candidates for regions that are affected by selection.
However, a robust inference depends on comparisons to a null distribution generated under a demographic model that must make
assumptions about human population history. The success of both strategies depends on the proximity of each marker to the target
of selection and the local recombination rate.

© 2003
Nature
Publishing
Group
110 | FEBRUARY 2003 | VOLUME 4 www.nature.com/reviews/genetics
REVIEWS
dependent,in part, on how well these signatures will be
able to predict the location of gene variants of biomed-
ical importance with little, if any, a priori knowledge of
their functional significance.To this end, the study of
population variation will continue to be of great interest
and relevance to researchers and clinicians.From this
process,we will learn more about the evolutionary his-
tory of our species,both the shared biology that makes
individuals so similar and the small fraction of differ-
ences that explain,in part,why one individual dies of
malaria, another is allergic to penicillin and another is
resistant to AIDS.
EPISTATIC
effects such as synergism or interference
between gene variants that lie close to one another
might also affect the patterns of selection in the
genome
110
.New multi-locus models of selection that
are designed to explore these effects seem promising
111
.
Our ability to interpret these patterns will improve as
we learn more about the signatures of selection at the
molecular level and as we improve our ability to link
causal variants to phenotypes. The clearest interpreta-
tions of the impact of selection are for loci about which
we already know a great deal.The increasing enthusiasm
for characterizing signatures of natural selection is
EPISTATIC
An interaction between non-
allelic genes, such that one gene
masks, interferes with or
enhances the expression of the
other gene.
1. Klein, R. G. The Human Career: Human Biological and
Cultural Origins (Univ. of Chicago Press, Chicago, 1999).
2. Klein, J. & Takahata, N. Where Do We Come From? The
Molecular Evidence for Human Descent (Springer, New
York, 2002).
3. Sober, E. The Nature of Selection: Evolutionary Theory in
Philosophical Focus (MIT Press, Cambridge,
Massachusetts, 1993).
4. Li, W. Molecular Evolution (Sinauer Associates, Sunderland,
Massachusetts, 1997).
An excellent introductory text that outlines the
theoretical basis of molecular evolutionary analyses
and provides insightful empirical examples.
5. Nei, M. Molecular Evolutionary Genetics (Columbia Univ.
Press, New York, 1987).
6. Endler, J. A. Natural Selection in the Wild (Princeton Univ.
Press, New Jersey, 1986).
7. Eyre-Walker, A. & Keightley, P. D. High genomic deleterious
mutation rates in hominids. Nature 397,344–347 (1999).
8. Kimura, M. Evolutionary rate at the molecular level. Nature
217, 624–626 (1968).
9. Kimura, M. Neutral Theory of Molecular Evolution
(Cambridge Univ. Press,Cambridge, UK, 1985).
10. Fay, J. C. & Wu, C. I. The neutral theory in the genomic era.
Curr. Opin. Genet. Dev. 11, 642–646 (2001).
11. Fay, J. C., Wyckoff, G. J. & Wu, C. I. Positive and negative
selection on the human genome. Genetics 158, 1227–1234
(2001).
12. Kreitman, M. Methods to detect selection in populations
with applications to the human. Annu. Rev. Genomics Hum.
Genet. 1, 539–559 (2000).
A detailed review of analytical methods to detect the
effects of natural selection on patterns of
polymorphism.
13. Fu, Y. X. Statistical tests of neutrality of mutations against
population growth, hitchhiking and background selection.
Genetics 147, 915–925 (1997).
14. Simonsen, K. L., Churchill, G. A. & Aquadro, C. F. Properties
of statistical tests of neutrality for DNA polymorphism data.
Genetics 141, 413–429 (1995).
15. Wall, J. D. Recombination and the power of statistical tests
of neutrality. Genet. Res. 74, 65–79 (1999).
16. Jorde, L. B., Watkins, W. S. & Bamshad, M. J. Human
population genomics: a bridge from evolutionary history to
genetic medicine. Mol. Genet. 10, 2199–2207 (2001).
17. Przeworski, M., Hudson, R. R. & Di Rienzo, A. Adjusting the
focus on human variation. Trends. Genet. 16, 296–302
(2000).
18. Ingman, M., Kaessmann, H., Paabo, S. & Gyllensten, U.
Mitochondrial genome variation and the origin of modern
humans. Nature408, 708–713 (2000).
19. Ke, Y. et al. African origin of modern humans in East Asia: a
tale of 12,000 Y chromosomes. Science292, 1151–1153
(2001).
20. Jorde, L. B. et al. The distribution of human genetic diversity:
a comparison of mitochondrial, autosomal and
Y-chromosome data. Am. J. Hum. Genet. 66, 979–988
(2000).
21. Kimmel, M. et al. Signatures of population expansion in
microsatellite repeat data. Genetics 148, 1921–1930
(1998).
22. Reich, D. E. & Goldstein, D. B. Genetic evidence for a
Paleolithic human population expansion in Africa. Proc. Natl
Acad. Sci. USA 95, 8119–8123 (1998).
23. Wooding, S. & Rogers, A. R. A Pleistocene population
X-plosion? Hum. Biol. 72, 693–695 (2000).
24. Rosenberg, N. A. et al. Genetic structure of human
populations. Science 298, 2381–2385 (2002).
The most comprehensive analysis of global patterns of
human population structure completed so far. It shows
that there is substantial geographical structure among
populations, although the proportion of an individual’s
ancestry from one or more of these populations is
highly variable.
25. Bamshad, M. et al. Human population genetic structure and
inference of group membership. Am. J. Hum. Genet. (in the
press).
26. Harpending, H. C. Genetic traces of ancient demography.
Proc. Natl Acad. Sci. USA 95, 1961–1967 (1995).
27. Takahata, N. Allelic genealogy and human evolution.
Mol. Biol. Evol. 10, 2–22 (1993).
28. Braverman, J. M., Hudson, R. R., Kaplan, N. L.,
Langley, C. H. & Stephan, W. The hitchhiking effect on the
site frequency spectrum of DNA polymorphisms. Genetics
140, 783–796 (1995).
29. Fullerton, S. M. et al. Geographic and haplotype structure of
candidate type 2 diabetes susceptibility variants at the
calpain-10 locus. Am. J. Hum. Genet. 70, 1096–1106
(2002).
30. Bamshad, M. J. et al. A strong signature of balancing
selection in the 5′cis-regulatory region of CCR5. Proc. Natl
Acad. Sci. USA 99, 10539–10544 (2002).
31. Stephens, J. C. et al. Haplotype variation and linkage
disequilibrium in 313 human genes. Science 293, 489–493
(2001).
An extensive survey of the level of polymorphism
found in and near more than 300 genes, which includes
a preliminary test of whether the patterns are
consistent with neutrality.
32. Prezworski, M. The signature of positive selection at
randomly chosen loci. Genetics 160,1179–1189 (2002).
33. Charlesworth, B. et al. The effect of deleterious mutations on
neutral molecular variation. Genetics 134, 1289–1303 (1993).
34. Hudson, R. R. & Kaplan, N. L. Deleterious background
selection with recombination. Genetics141, 1605–1617
(1995).
35. Maynard-Smith, J. & Haigh, J. The hitch-hiking effect of a
favorable gene. Genet. Res. 23, 23–35 (1974).
A noteworthy exposition of the impact of positive
selection on linked neutral polymorphisms.
36. Kaplan, N. L., Hudson, R. R. & Langley, C. H. The ‘hitchhiking
effect’ revisited. Genetics 123,887–899 (1989).
37. Begun, J. J. & Aquadro, C. F. Levels of naturally occurring
DNA polymorphism correlate with recombination rates in
D. melanogaster. Nature356, 519–520 (1992).
38. Nachman, M. W. Patterns of DNA variability at X-linked loci in
Mus domesticus. Genetics147, 1303–1316 (1997).
39. Nachman, M. W. Single nucleotide polymorphisms and
recombination rate in humans. Trends Genet. 17, 481–485
(2001).
40. Kim, Y. & Stephan, W. Joint effects of genetic hitchhiking and
background selection on neutral variation. Genetics155,
1415–1427 (2000).
41. Fay, J. C., Wyckoff, G. J. & Wu, C. I. Testing the neutral theory
of molecular evolution with genomic data from Drosophila.
Nature 415, 1024–1026 (2002).
42. Yang, Z. Inference of selection from multiple species
alignments. Curr. Opin. Genet. Dev. 12, 1–7 (2002).
43. Bush, R. M. Predicting adaptive evolution. Nature Rev.
Genet. 2, 387–392 (2001).
44. Wyckoff, G. J., Wang, W. & Wu, C. I. Rapid evolution of male
reproductive genes in the descent of man. Nature403,
304–309 (2000).
45. Johnson, M. E. et al. Positive selection of a gene family during
the emergence of humans and African apes. Nature 413,
514–519 (2001).
46. Enard, W. et al. Molecular evolution of FOXP2, a gene
involved in speech and language. Nature 418, 869–872
(2002).
47. Olsen, M. V. & Varki, A. Sequencing the chimpanzee
genome: insights into human evolution and disease.
Genetics 4, 20–28 (2003).
48. Neel, J. V. Diabetes mellitus: a ‘thrifty’ genotype rendered
detrimental by ‘progress’? Am. J. Hum. Genet. 14,
353–362 (1962).
49. Wooding, S. P. et al. DNA sequence variation in a 3.7-kb
noncoding sequence 5′ of the CYP1A2 gene: implications
for human population history and natural selection. Am. J.
Hum. Genet. 71, 528–542 (2002).
50. Sabeti, P. C. et al. Detecting recent positive selection in the
human genome from haplotype structure. Nature 419,
832–837 (2002).
51. Tishkoff, S. A. et al. Haplotype diversity and linkage
disequilibrium at human G6PD: recent origin of alleles that
confer malarial resistance. Science 293, 455–462 (2001).
52. Saunders, M. A., Hammer, M. F. & Nachman M. W.
Nucleotide variability at G6PD and the signature of malarial
selection in humans. Genetics (in the press).
53. Toomajian, C. & Kreitman, M. Sequence variation and
haplotype structure at the human HFE locus. Genetics 161,
1609–1623 (2002).
54. Harding, R. M. Archaic African and Asian lineages in the
genetic ancestry of modern humans. Am. J. Hum. Genet.
70, 369–383 (1997).
55. Wooding, S. & Rogers, A. The matrix coalescent and an
application to human single-nucleotide polymorphisms.
Genetics 161, 1641–1650 (2002).
56. Hamblin, M. T. & Di Rienzo, A. Detection of the signature of
natural selection in humans: evidence from the Duffy blood
group locus. Am. J. Hum. Genet. 66, 1669–1679 (2000).
57. Hamblin, M. T., Thompson, E. E. & Di Rienzo, A. Complex
signatures of natural selection at the Duffy blood group
locus. Am. J. Hum. Genet. 70, 369–383 (2002).
A meticulous analysis of the molecular signature of
selection on a classical human trait, which illustrates
the potential confounding effects of population
history and the interaction of several selective forces.
58. Harding, R. M. et al. Evidence for variable selective
pressures at MC1R. Am. J. Hum. Genet. 66, 1351–1361
(2000).
59. Makova, K. D., Ramsay, M., Jenkins, T. & Li, W. H. Human
DNA sequence variation in a 6.6-kb region containing the
melanocortin 1 receptor promoter. Genetics 158,
1253–1268 (2001).
60. Ding, Y. C. et al. Evidence of positive selection acting at the
human dopamine receptor D4 gene locus. Proc. Natl Acad.
Sci. USA 99, 309–314 (2002).
61. Harris, E. E. & Hey, J. Human populations show reduced
DNA sequence variation at the factor IX locus. Curr. Biol.
11, 774–778 (2001).
62. Nachman, M. W. & Crowell, S. L. Contrasting evolutionary
histories of two introns of the Duchenne muscular
dystrophy gene, Dmd, in humans. Genetics 155,
1855–1864 (2000).
63. Gilad, Y., Rosenberg, S., Przeworski, M., Lancet, D. &
Skorecki, K. Evidence for positive selection and population
structure at the human MAO-A gene. Proc. Natl Acad. Sci.
USA 99,862–867 (2002).
64. Enattah, N. S. et al. Identification of a variant associated
with adult-type hypolactasia. Nature Genet. 30, 233–237
(2002).
A good example of the difficulties of finding the
functional variants under selection at a locus with a
signature of positive selection.

© 2003
Nature
Publishing
Group
NATURE REVIEWS | GENETICS VOLUME 4 | FEBRUARY 2003 | 111
REVIEWS
65. Stephens, J. C. et al. Dating the origin of the CCR5-
∆
32
AIDS-resistance allele by the coalescence of haplotypes.
Am. J. Hum. Genet. 62, 1507–1515 (1998).
66. Leber, F. et al. The ∆32-ccr5 mutation conferring protection
against HIV-1 in Caucasian populations has a single and
recent origin in Northeastern Europe. Hum. Mol. Genet. 7,
399–406 (1998).
67. Roses, A. D. Pharmacogenetics and the practice of
medicine. Nature405, 857–865 (2001).
68. Scordo, M. G. & Spina M. Cytochrome P450
polymorphisms and response to antipsychotic therapy.
Pharmacogenomics 31, 1–18 (2002).
69. Ikeya, K. et al. Human CYP1A2: sequence, gene structure,
comparison with the mouse and rat orthologous gene, and
differences in liver 1A2 mRNA expression. Mol. Endocrinol.
3, 1399–1408 (1989).
70. Rosenberg, N. A. & Nordborg, M. Genealogical trees,
coalescent theory and the analysis of genetic
polymorphisms. Nature Rev. Genet. 3, 380–390 (2002).
71. Hudson, R. R. & Kaplan, N. L. The coalescent process in
models with selection and recombination. Genetics120,
831–840 (1988).
72. Shi, Y., Radlwimmer, F. B. & Yokoyama, S. Molecular
genetics and the evolution of ultraviolet vision in vertebrates.
Proc. Natl Acad. Sci. USA 98, 11731–11736 (2001).
73. Nordborg, M. & Tavare, S. Linkage disequilibrium: what
history has to tell us. Trends Genet. 18, 83–90 (2002).
74. Livingstone, F. B. Malaria and human polymorphisms.
Annu. Rev. Genet. 5, 33–64 (1974).
75. Cooke, G. S. & Hill, A. V. S. Genetics of susceptibility to
human infectious disease. Nature Rev. Genet. 2, 967–977
(2001).
76. Miller, L. H. Impact of malaria on genetic polymorphism and
genetic diseases in Africans and African Americans.
Proc. Natl Acad. Sci. USA 91, 2415–2419 (1974).
77. Vulliamy, T. J., Mason, P. & Luzzatto, L. The molecular
basis of glucose-6-phosphate dehydrogenase deficiency.
Trends Genet. 8, 138–142 (1992).
78. Beutler, E. G6PD deficiency. Blood 84, 3613–3636 (1994).
79. Ruwende, C. et al. Natural selection of hemi- and
heterozygotes for G6PD deficiency in Africa by resistance to
severe malaria. Nature376, 246–249 (1995).
80. Austin, L. H. & Federica, V. Very large long-term effective
population size in the virulent human malaria parasite
Plasmodium falciparum. Proc. R. Soc. Lond. B 268,
1855–1860 (2001).
81. Coluzzi, M. The clay feet of the malaria giant and its African
roots: hypotheses and inferences about origin, spread and
control of Plasmodium falciparum. Parassitologia 41,
277–283 (1999).
82. Feder, J. N. et al. A novel MHC class I-like gene is mutated
in patients with hereditary haemochromatosis. Nature
Genet. 13, 399–408 (1996).
83. Merryweather-Clarke, A. T., Pointon, J. J., Shearman, J. D.
& Robson, K. J. Global prevalence of putative
haemochromatosis mutations. J. Med. Genet. 34, 275–278
(1997).
84. Ajioka, R. S. et al. Haplotype analysis of hemochromatosis:
evaluation of different linkage-disequilibrium approaches
and evolution of disease chromosomes. Am. J. Hum.
Genet. 60, 1439–1447 (1997).
85. Thomas, W. et al. Haplotype and linkage disequilibrium
analysis of the hereditary hemochromatosis gene region.
Hum. Genet. 102, 517–525 (1998).
86. Bulaj, Z. J., Griffen, L. M., Jorde, L. B., Edwards, C. Q. &
Kushner, J. P. Clinical and biochemical abnormalities in
people heterozygous for hemochromatosis. N. Engl. J. Med.
335, 1799–1805 (1996).
87. Scholl, T. O., Hediger, M. L., Fischer, R. L. & Shearer, J. W.
Anemia vs. iron deficiency: increased risk of preterm delivery
in a prospective study. Am. J. Clin. Nutr. 55, 985–988
(1992).
88. Pritchard, J. K. Are rare variants responsible for susceptibility
to complex diseases? Am. J. Hum. Genet. 69, 124–137
(2001).
89. Hawkes, K., O’Connell, J. F., Blurton Jones, N. G.,
Alvarez, H. & Charnov, E. L. Grandmothering, menopause,
and the evolution of human life histories. Proc. Natl Acad.
Sci. USA 95, 1336–1339 (1998).
90. Lewontin, R. C. & Hubby, J. L. A molecular approach to
the study of genetic heterozygosity in natural populations.
II. Amount of variation and degree of heterozygosity in
natural populations of Drosophila pseudoobscura.
Genetics 54, 595–609 (1966).
91. Kaplan, N. L., Darden, T. & Hudson, R. R. The coalescent
process in models with selection. Genetics 120, 819–829
(1988).
92. Richman, A. D. & Kohn, J. R. Self-incompatibility alleles
from Physalis: implications for historical inference from
balanced genetic polymorphisms. Proc. Natl Acad. Sci.
USA 96, 168–172 (1999).
93. Hughes, A. L. & Yeager, M. Natural selection at major
histocompatibility complex loci of vertebrates. Annu. Rev.
Genet. 32, 415–435 (1998).
94. Verrelli, B. C. et al. Evidence for balancing selection from
nucleotide sequence analyses of human G6PD. Am. J.
Hum. Genet. 71, 1112–1128 (2002).
95. Baum, J., Ward, R. H. & Conway, D. H. Natural selection
on the erythrocyte surface. Mol. Biol. Evol. 19, 223–229
(2002).
96. Wu, X., Di Rienzo, A. & Ober, C. A population genetics
study of single nucleotide polymorphisms in the interleukin
4 receptor a (IL4RA) gene. Genes Immun. 2, 128–134
(2001).
97. Liu, R. et al. Homozygous defect in HIV-1 coreceptor
accounts for resistance of some multiply exposed
individuals to HIV-1 infection. Cell 86, 367–377
(1996).
98. Schierup, M. H., Vekemans, X. & Charlesworth, D. The
effect of subdivision on variation at multi-allelic loci under
balancing selection. Genet. Res. 76,51–62 (2000).
99. Charlesworth, B., Nordborg, M. & Charlesworth, D. The
effects of local selection, balanced polymorphism and
background selection on equilibrium patterns of genetic
diversity in subdivided populations. Genet. Res. 70,
155–174 (1997).
100. Takahata, N. & Nei, M. Allelic genealogy under
overdominant and frequency-dependent selection and
polymorphism of major histocompatibility complex loci.
Genetics 124, 967–978 (1990).
101. Salamon, H. et al. Evolution of HLA class II molecules:
allelic and amino acid site variability across populations.
Genetics 152, 393–400 (1999).
102. Grimsley, C., Mather, K. A. & Ober, C. HLA-H: a
pseudogene with increased variation due to balancing
selection at neighboring loci. Mol. Biol. Evol. 15,
1581–1588 (1998).
103. Muller, H. J. Our load of mutation. Am. J. Hum. Genet. 2,
111–176 (1950).
104. Gillespie, J. H. The Causes of Molecular Evolution (Oxford
Univ. Press, New York, 1991).
105. Payseur, B. A., Cutter, A. D. & Nachman, M. W. Searching
for evidence of positive selection in the human genome
using patterns of microsatellite variability. Mol. Biol. Evol.
19, 1143–1153 (2002).
106. Cargill, M. et al. Characterization of single-nucleotide
polymorphisms in coding regions of human genes. Nature
Genet. 22, 231–238 (1999).
107. Sunyaev, S. R., Lathe, W. C., Ramensky, V. E. & Bork, P.
SNP frequencies in human genes an excess of rare alleles
and differing modes of selection. Trends Genet. 16,
335–337 (2000).
108. Akey, J. M., Zhang, G., Zhang, K., Jin, L. & Shriver, M. D.
Interrogating a high-density SNP map for signatures of
natural selection. Genome Res. 12, 1805–1814 (2002).
109. Wiehe, T. The effect of selective sweeps on the variance
of the allele distribution of a linked multiallele locus:
hitchhiking of microsatellites. Theor. Popul. Biol. 53,
272–283 (1998).
110. Comeron, J. M. & Kreitman, M. Population, evolutionary
and genomic consequences of interference selection.
Genetics 161, 389–410 (2002).
111. Navarro, A. & Barton, N. H. The effects of multilocus
balancing selection on neutral variability. Genetics 161,
849–863 (2002).
112. Watterson, G. A. On the number of segregating sites in
genetical models without recombination. Theor. Popul.
Biol. 7, 256–276 (1975).
113. Tajima, F. Evolutionary relationships of DNA sequences in
finite populations. Genetics 105, 437–460 (1983).
114. Sachidanandam, R. et al. A map of human genome
sequence variation containing 1.42 million single nucleotide
polymorphisms. Nature409, 928–933 (2001).
115. Li, W. H. & Sadler, L. A. Low nucleotide diversity in man.
Genetics 129, 513–523 (1991).
116. Reich, D. E. et al. Human genome sequence variation and
the influence of gene history, mutation and recombination.
Nature Genet. 32, 135–142 (2002).
117. Zietkiewicz, E. et al. Genetic structure of the ancestral
population of modern humans. J. Mol. Evol. 47, 146–155
(1998).
118. Tajima, F. Statistical method for testing the neutral mutation
hypothesis by DNA polymorphism. Genetics 123,585–595
(1989).
119. Fay, J. C. & Wu, C. I. Hitchhiking under positive Darwinian
selection. Genetics155, 1405–1413 (2000).
Introduces a new statistical test of neutrality on the
basis of the prediction that, immediately after a
selective sweep, an excess of high-frequency-derived
polymorphisms is expected at linked sites.
120. Przeworski, M. The signature of positive selection at
randomly chosen loci. Genetics 160, 1179–1189 (2002).
121. Gilad, Y. et al. Dichotomy of single-nucleotide polymorphism
haplotypes in olfactory receptor genes and pseudogenes.
Nature Genet. 26, 221–224 (2000).
122. Huttley, G. A. et al. Adaptive evolution of the tumor
suppressor BRCA1 in humans and chimpanzees. Nature
Genet. 25, 410–413 (2000).
123. Suzuki, Y. & Gojobori, T. A method for detecting positive
selection at single amino acid sites. Mol. Biol. Evol. 16,
1315–1328 (1999).
124. Schlotterer, C. Towards a molecular characterization of
adaptation in local populations. Curr. Opin. Genet. Dev. 12,
1–4 (2002).
125. Lewontin, R. C. & Krakauer, J. Distribution of gene
frequency as a test of the theory of the selective neutrality of
polymorphisms. Genetics74, 175–195 (1973).
126. Bowcock, A. M. et al. Drift, admixture, and selection in
human evolution: a study with DNA polymorphisms.
Proc. Natl Acad. Sci. USA 88, 839–843 (1991).
127. Beaumont, M. A. & Nichols, R. A. Evaluating loci for use in
genetic analysis of population structure. Proc. R. Soc. Lond.
B 263, 1619–1626 (1996).
128. McDonald, J. H. & Kreitman, M. Adaptive protein evolution
at the ADH locus in Drosophila. Nature 351, 652–654
(1991).
129. Fu, X. Y. & Li, W. H. Statistical tests of neutrality of mutations.
Genetics 133, 693–709 (1993).
130. Li, W. H., Wu, C. I. & Luo, C. C. A new method for
estimating synonymous and nonsynonymous rates of
nucleotide substitution considering the relative likelihood of
nucleotide and codon changes. Mol. Biol. Evol. 2, 150–174
(1985).
131. Nei, M. & Gojobori, T. Simple methods for estimating the
numbers of synonymous and nonsynonymous nucleotide
substitutions. Mol. Biol. Evol. 3, 418–426 (1985).
132. Hudson, R. R., Kreitman, M. & Aguade, M. A test of neutral
molecular evolution based on nucleotide data. Genetics
116, 153–159 (1987).
Acknowledgements
We thank L. B. Jorde, A. R. Rogers and three anonymous review-
ers for comments and criticisms. The authors are supported by
funds from the US National Institutes of Health and the National
Science Foundation.
Online links
DATABASES
The following terms in this article are linked online to:
LocusLink: http://www.ncbi.nlm.nih.gov/LocusLink
BRCA1 | calpain-10 | CCR5 | CYP1A2 | DRD4| G6PD | MAOA |
MC1R | PPARG
OMIM: http://www.ncbi.nlm.nih.gov/Omim
idiopathic haemochromatosis | type II diabetes
Access to this interactive links box is free online.