Content uploaded by Sagiv Shifman
Author content
All content in this area was uploaded by Sagiv Shifman on Oct 17, 2014
Content may be subject to copyright.
NEWS AND VIEWS
118 VOLUME 37
|
NUMBER 2
|
FEBRUARY 2005
|
NATURE GENETICS
to phosphorylate itself and its downstream
targets, and this could be part of its mecha-
nism for inhibiting the activation of DNA-
repair complexes both at telomeres and at
double-strand breaks. The demonstration by
Bradshaw et al. that TRF2 is rapidly recruited
to generic double-strand breaks will initiate
a mutually productive period of interaction
between the fields of DNA repair and telo-
mere biology, as the roles for telomeric fac-
tors in the choreography of repair come into
the spotlight.
1. Bradshaw, P.S., Stavropoulos, D.J. & Meyn, M.S. Nat.
Genet. 37, 193–197 (2005).
2. de Lange, T. Nat. Rev. Mol. Cell. Biol. 5, 323–329
(2004).
3. Griffith, J.D. et al. Cell 97, 503–514 (1999).
4. Stansel, R.M., de Lange, T. & Griffith, J.D. EMBO J.
20, 5532–5540 (2001).
5. Hardy, C.F., Sussel, L. & Shore, D. Genes Dev. 6,
801–814 (1992).
6. Silverman, J., Takai, H., Buonomo, S.B., Eisenhaber,
F. & de Lange, T. Genes Dev. 18, 2108–2119
(2004).
7. Xu, L. & Blackburn, E.H. J. Cell. Biol. 167, 819–830
(2004).
8. van Steensel, B., Smogorzewska, A. & de Lange, T.
Cell 92, 401–413 (1998).
9. Wang, R.C., Smogorzewska, A. & de Lange, T. Cell
119, 355–368 (2004).
10. Gommers-Ampt, J., Lutgerink, J. & Borst, P. Nucleic
Acids Res. 19, 1745–1751 (1991).
11. Steinert, S., Shay, J.W. & Wright, W.E. Mol. Cell. Biol.
24, 4571–4580 (2004).
12. Karlseder, J. et al. PLoS Biol. 2, E240 (2004)
The beauty of admixture
Ariel Darvasi & Sagiv Shifman
Admixture mapping is an old concept that has only now been applied with markers across the entire genome.
Such a study scanning an African American population identified two chromosomal regions affecting
susceptibility to hypertension.
Anecdotally, children of parents of mixed
ethnicities are exotically beautiful. More sci-
entifically established is the merit of admixed
populations for gene mapping purposes. The
potential value of admixed populations was
suggested more than half a century ago
1
.
Substantial theoretical and practical aspects
have been developed since then (reviewed
by McKeigue
2
). A genome scan to identify
genes affecting a complex trait is now pre-
sented for the first time to our knowledge
by Xiaofeng Zhu and colleagues on page 177
of this issue
3
.
The admixed population
The concept behind admixture mapping is
simple (Fig. 1). In essence, admixture map-
ping is most similar to linkage analysis in
experimental crosses with inbred strains,
with specific similarity to advanced inter-
cross lines
4
. An advanced intercross line is a
population derived from two inbred strains
that were randomly intercrossed for several
generations. An advanced intercross line
constitutes the ideal admixed population: all
variations can be identified in one of the two
progenitors, the mean ancestral composition
is 50% for each progenitor, allele frequencies
in the progenitor populations are either 1 or
0, and random mating is followed after a sin-
gle generation of intercrossing the progeni-
tors. In a human admixed population, these
ideal conditions will never be met, resulting
in decreased power for mapping purposes.
Except for gene effect, which has a strong
influence on power, the parameter that mostly
affects power, specifically in admixture map-
ping, is the extent of difference in allele fre-
quency between the ancestral populations
5
.
Ariel Darvasi is in The Life Sciences Institute,
The Hebrew University, Jerusalem 91904, Israel.
Sagiv Shifman is in the Wellcome Trust Centre
for Human Genetics, Oxford OX3 7BN, UK. e-
mail: arield@cc.huji.ac.il, sagiv@well.ox.ac.uk
Case Control
Population 1
Population 2
Disease gene location
Figure 1 Schematic of one chromosome pair from each of several individuals in an admixed population.
A group of cases (for a given disease) and a group of controls are separately presented at the bottom left
and the bottom right, respectively. For one of the control individuals (arrow), a schematic presentation of
all its ancestors in the last four generations is shown in the upper part of the figure. Admixture mapping
can be ideally applied if population 1 (blue) and population 2 (red) carry a different allele at the disease
locus (dashed line). Whole-genome scanning under the admixture mapping strategy consists of scanning
the genome and identifying the regions with an excess of ‘red’ ancestry in the cases versus the controls,
assuming that the ‘red’ population carries the predisposition allele. The size of the blocks from different
ancestors will depend on the number of generations since the populations were mixed.
© 2005 Nature Publishing Group http://www.nature.com/naturegenetics
NEWS AND VIEWS
NATURE GENETICS
|
VOLUME 37
|
NUMBER 2
|
FEBRUARY 2005 119
For example, in the extreme case where the
allele affecting a disease has the same fre-
quency in both ancestral populations, admix-
ture mapping cannot be efficiently applied. In
contrast, the power of admixture mapping
will be only mildly affected by the percentage
contributed by each population to the admix-
ture, as long as that proportion is between
20% and 80% (ref. 5).
Genome scan for hypertension
In an effort to identify chromosomal regions
affecting hypertension, Zhu et al.
3
carried out
a genome scan with 269 microsatellite mark-
ers and a total of 737 cases (hypertensives)
and 573 controls (normotensives). Cases
and controls were selected from the African
American population. African Americans are
an admixed population with ∼75% African
ancestry and ∼25% European ancestry
6
and
are thus appropriate for admixture mapping.
All individuals were sampled from three net-
works (GenNet, GENOA and HyperGEN) in
geographically distinct locations participat-
ing in the Family Blood Pressure Program.
Zhu et al.
3
initially explored hyperten-
sive cases only, independently in the three
networks, and found an excess of African
ancestry in more than one network on chro-
mosomes 4, 6 and 21. In particular, two
markers around 6q24 showed an excess of
African ancestry in all three populations.
To validate the significance of these results,
they compared the excess of African ances-
try found in the cases with that found in
controls. The excess of African ancestry was
shifted upwards in cases relative to controls.
The entire shift can be attributed to two chro-
mosomal regions at 6q24 and 21q21 where
the excess of African ancestry was signifi-
cant in cases but not in controls. Therefore,
these findings suggest that the chromosomal
regions 6q24 and 21q21 contain genes affect-
ing predisposition to hypertension. Support
for the chromosome 6q24 findings can be
drawn from previous linkage studies that
found evidence for linkage between this
chromosomal region and hypertension
or related traits
7,8
. The large size of this
chromosomal region (37 cM, including all
markers with Z score >2.5) may suggest that
more than one gene affecting hypertension is
present. This is not unexpected, as cis-acting
linked genes will behave as a single gene with
a larger effect (the combined effect of the two
genes) in an admixture mapping experiment,
hence having greater power of being picked
up in a genome scan. The 21q21 region needs
further replication to establish its validity, as
this region has not previously been suggested
to be associated with hypertension.
A complementary approach
Two main approaches have been used to
search for genes affecting complex traits:
linkage analysis and association analysis
9
.
Linkage analysis has two key disadvantages:
relatively low statistical power for detecting
modest effects
10
, and low mapping resolution,
which prevents gene identification even after
a region has been detected
9
. Association anal-
ysis also has two key disadvantages. Because
this approach is based on linkage disequilib-
rium or on testing the potential functional
polymorphisms, the number of polymor-
phisms that need to be scanned in the entire
genome is painfully high (>100,000)
11
. The
second disadvantage is the diminishing power
that occurs with high genetic heterogene-
ity
12
. Admixture mapping is a strategy that
falls between linkage analysis and association
analysis in many respects (Table 1).
Although admixture mapping has a sub-
stantially lower mapping resolution than
association analysis, as long as genotyping
costs are a limiting factor, admixture map-
ping will be a good approach for the initial
genome scan. Admixture mapping is particu-
larly appropriate for traits for which there is a
large difference in the phenotypic prevalence
in the ancestral populations of the admix-
ture. Nevertheless, admixture mapping is
not limited to those traits and will still work
if the allele frequencies of the disease locus
are different in the ancestors of the admixed
population. This is more likely to occur when
the disease prevalence varies in the ancestral
populations.
Given the advantages of admixture map-
ping, it is notable that this experiment has
only now been done. One reason for this
might be the notion (which might be cor-
rect) that more markers are required for an
adequate whole-genome scan with admix-
ture mapping
5
than were used in the current
experiment. In addition, admixture mapping
is efficient only if the allele frequencies of
the markers are substantially different in the
ancestral populations. In that respect, it now
seems that microsatellite panels might be
more informative than originally thought
13
.
Consequently, a standard panel of markers,
normally used in linkage experiments, suc-
cessfully served Zhu et al.
3
in their admix-
ture mapping study. A word of caution is
appropriate, though. The unexpected success
might be due to the specific constellations
particular to the current experiment, includ-
ing chance. Therefore, the study of Zhu et
al.
3
, which applied admixture mapping to
hypertension and concluded with success-
ful and robust results, will still require some
replications in other traits and with other
samples before its generality can be estab-
lished. The current results, however, are
undoubtedly promising enough to encour-
age the scientific community to carry out
these essential replications.
1. Rife, D.C. Am. J. Hum. Genet. 6, 26–33 (1954).
2. McKeigue, P.M. Am. J. Hum. Genet. 76, 1–7
(2005).
3. Zhu, X. et al. Nat. Genet. 37, 177–181 (2005).
4. Darvasi, A. & Soller, M. Genetics 141, 1199–1207
(1995).
5. Patterson, N. et al. Am. J. Hum. Genet. 74, 979–
1000 (2004).
6. Destro-Bisol, G. et al. Hum. Genet. 104, 149–157
(1999).
7. Krushkal, J. et al. Circulation 99, 1407–1410
(1999).
8. Arya, R. et al. Diabetes 51, 841–847 (2002).
9. Lander, E.S. & Schork, N.J. Science 265, 2037–
2048 (1994).
10. Risch, N. & Merikangas, K. Science 273, 1516–1517
(1996).
11. Risch, N.J. Nature 405, 847–856 (2000).
12. Weiss, K.M. & Terwilliger, J.D. Nat Genet. 26, 151–
157 (2000).
13. Tang, H. et al. Am. J. Hum. Genet. (in the press).
Table 1 Main characteristics of mapping strategies
Linkage analysis Admixture mapping Association analysis
Statistical power Low High
*
High
Number of SNPs required for whole
genome scan
Low Low High
Sensitivity to genetic heterogeneity Low Moderate High
Mapping resolution Poor Intermediate Good
*Power diminishes to zero with equal allele frequencies in the ancestral population.
© 2005 Nature Publishing Group http://www.nature.com/naturegenetics