Genome-wide linkage analysis of 723 affected
relative pairs with late-onset Alzheimer’s disease
Marian L. Hamshere1,2, Peter A. Holmans1,2, Dimitrios Avramopoulos3,4, Susan S. Bassett3,
Deborah Blacker5,7, Lars Bertram6, Howard Wiener8, Nan Rochberg9, Rudolph E. Tanzi6,
Amanda Myers10, Fabienne Wavrant-De Vrie `ze11, Rodney Go8, Daniele Fallin3,
Simon Lovestone12, John Hardy13, Alison Goate9, Michael O’Donovan2, Julie Williams1,2
and Michael J. Owen2,*
1Biostatistics and Bioinformatics Unit, and2Department of Psychological Medicine, School of Medicine, Cardiff
University, Heath Park, Cardiff CF14 4XN, UK,3Department of Psychiatry, and4McKusick Nathans Institute of
Genetic Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, USA,5Gerontology Research Unit,
Department of Psychiatry, and6Genetics and Aging Research Unit, Department of Neurology, Massachusetts
General Hospital, Harvard Medical School, Charlestown, MA 02129, USA,7Department of Epidemiology, Harvard
School of Public Health, Boston, MA 02115, USA,8Department of Epidemiology, University of Alabama at
Birmingham, USA,9Department of Psychiatry, Washington University School of Medicine, 660 S. Euclid Avenue,
St Louis, MO 63110, USA,10Department of Psychiatry and Behavioral Sciences, Miller School of Medicine, University
of Miami, USA,11Laboratory of Neurogenetics, National Institute on Aging, National Institutes of Health, 35 Convent
Drive, Bethesda, MD 20892-3707, USA,12MRC Centre for Neurodegeneration Research, King’s College London,
Institute of Psychiatry, De Crespigny Park, London SE5 8AF, UK and13Department of Molecular Neuroscience and
Reta Lila Weston Laboratories, Institute of Neurology, UCL, Queen Square, London WC1N 3BG, UK
Received August 11, 2007; Revised and Accepted August 12, 2007
Previous attempts to identify genetic loci conferring risk for late-onset Alzheimer’s disease (LOAD) through
linkage analysis have observed some regions of linkage in common. However, due to the sometimes-
considerable overlap between the samples, some of these reports cannot be considered to be independent
replications. In order to assess the strength of the evidence for linkage and to obtain the best indication of
the location of susceptibility genes, we have amalgamated three large samples to give a total of 723 affected
relative pairs (ARPs). Multipoint, model-free ARP linkage analysis was performed. Genome-wide significant
evidence for linkage was observed on 10q21.2 (LOD 5 3.3) and genome-wide suggestive evidence was
observed on 9q22.33 (LOD 5 2.5) and 19q13.32 (LOD 5 2.0). One further region on 9p21.3 was identified
with an LOD score > 1. We observe no evidence to suggest that more than one locus is responsible for
the linkage to 10q21.2, although this linked region may harbour more than one susceptibility gene.
Evidence of allele-sharing heterogeneity between the original collection sites was observed on chromosome
9 but not on chromosome 10 or 19. Evidence for an interaction was observed between loci on chromosomes
10 and 19. Where samples overlapped, the genotyping consistency was high, estimated to average at 97.3%.
Our large-scale linkage analysis consolidates clear evidence for a susceptibility locus for LOAD on 10q21.2.
Alzheimer’s disease (AD) is a heritable (1), debilitating dis-
order characterized by a gradual decline in cognitive abilities.
AD is estimated to account for two-thirds of individuals with
dementia (2) and to affect approximately 15 million people
worldwide (3). Variants of three genes (APP, PS-1 and
PS-2) play a major role in the genetics of early-onset autoso-
mal dominant AD (4–6). The great majority of AD cases are
of late onset and show complex, non-Mendelian patterns of
# The Author 2007. Published by Oxford University Press. All rights reserved.
For Permissions, please email: firstname.lastname@example.org
*To whom correspondence should be addressed. Tel: þ44 2920743058; Fax: þ44 2920746554; Email: email@example.com
Human Molecular Genetics, 2007, Vol. 16, No. 22
Advance Access published on August 27, 2007
by guest on June 7, 2013
inheritance. Late-onset AD (LOAD) probably results from the
combined effects of variations in a number of genes as well as
environmental factors. Recent twin studies suggest that the
heritability of AD is in the range of 60–80%, with the
balance of variance attributable to unique (typically adult)
environmental influences (1,7). The only genetic risk factor
for LOAD to have been identified with certainty is the apoli-
poprotein E (APOE) gene. The APOE4 allele is highly
enriched in AD cases compared with non-demented individ-
uals, whereas the APOE2 allele is under-represented in cases
(8). However, only 50% of AD cases carry an APOE4 allele
(9–11), and it is clear that other genes play a role, in particular
to the age at onset of illness (12). Many other positive findings
have been reported from association studies of functional and
positional candidate genes, but unequivocal evidence for
association has yet to be obtained. A database presenting
meta-analyses of current candidate gene association studies
can be found on the Alzheimer Research Forum at
Linkage analysis normally provides the first step in identify-
ing regions of the genome that could harbour novel suscepti-
bility genes. There have been a number of genome-wide
linkage studies of AD using large samples of families with
two or more affected relatives (14–21). The major regions
identified are found on chromosomes 9p, 9q, 12p and 19q,
with the most convincing evidence for linkage on 10q
(14,16,22–25). These results support the possibility that a
AD risk and which remain to be identified. However, many of
the family samples used in the linkage analyses overlap, and
this has made it difficult to assess the strength and consistency
of the evidence in favour of linkage. We have therefore com-
bined data from a number of the published linkage studies,
taking full account of overlap. The complex two-stage design
of one of the studies (14,16) resulted in the total number of
affected relative pairs (ARPs) with markers genotyped for
different chromosomal regions varying between 423 and 710
in the total sample of 723 ARPs, and the largest number of
available ARPs genotyped was for markers on chromosome
10. Given the mounting evidence to support an AD suscepti-
bility gene on chromosome 10, we were particularly keen to
analyse this chromosome to obtain the best evidence for
linkage and the most accurate estimate of gene location from
the available data since previous studies have provided discre-
pant estimates of the location of the putative risk gene. Given
that many markers were genotyped more than once in different
laboratories, combining the data also allowed us to assess geno-
typing quality between research groups.
Four regions produced a maximum LOD.1 on chromosomes
9p21.3, 9q22.33, 10q21.2 and 19q13.32 (Table 1 and Fig. 1).
In addition, all four regions were identified as being of interest
when the genotypes from the three groups were analysed sep-
arately (see Supplementary Material, Table S1 for more
details) and showed no evidence of allele heterogeneity.
Chromosomes 9 and 10 showed no significant increase in
maximum LOD when APOE was included as a covariate in
the model, whereas chromosome 19 showed a potential age
at onset effect.
The highest LOD score was observed on chromosome 10 with
a value of 3.3 at 78 cM (nearest marker ¼ D10S464), with an
Table 1. All maximum LOD scores .1 in the full sample and their relevant
Genetic location of
maximum LOD (cM)
D9S1870 and D9S741
Figure 1. Plots of all chromosomes where the maximum LOD.1 in the full
2704Human Molecular Genetics, 2007, Vol. 16, No. 22
by guest on June 7, 2013
identity by descent (IBD) estimate of 0.58. This maximum
LOD score equals the criterion required for significant
genome-wide linkage (26). The LOD-1 region spans 17 cM
(between 68 and 85 cM). The maximum LOD score is
higher than the LOD scores obtained when analysing data
from the three groups separately; note, however, that in their
second stage linkage analysis, group 1 (16) had reported a
maximum LOD score of 3.9 near 82 cM. The maximum
LOD score for the sample studied by group 2 lies at 88 cM,
which is outside the overall LOD-1 region, suggesting that
inclusion of the additional data from groups 1 and 3 has pro-
vided extra information. A test of homogeneity in allele
sharing between sites of collection was carried out by incor-
porating site as a 5-level factor in a covariate linkage analysis
(see Method of Analysis). This gave a chromosome-wide
P-value of 0.18, indicating no significant differences in
allele sharing between the five sites. However, the power of
such a test may be low due to the multiple degrees of
freedom incurred by testing differences between all sites sim-
ultaneously. The location of the maximum LOD score when
the site covariate was included was 62 cM (outside the
LOD-1 region) and the estimates of IBD allele-sharing prob-
abilities (compared with an expected value of 0.5 in the
absence of linkage) were 0.51, 0.46, 0.56, 0.60 and 0.58 in
the UK, NIMH (National Institute of Mental Health Genetics
Initiative for Alzheimer’s Disease) sites 50, 51 and 52 and the
NIA (National Institute on Aging, National Cell Repository
for Alzheimer’s Disease) samples, respectively. Thus, there
may be some heterogeneity of IBD sharing between sites
despite the overall test for homogeneity being non-significant.
The analysis described in the previous paragraph does not
test for homogeneity in the location of the maximum LOD
score in each sample (which may indicate the presence of
two or more disease-susceptibility genes in the region). For
a complex trait like LOAD, testing for such homogeneity in
the whole sample would be complicated by the likely presence
of several pedigrees that do not segregate any disease gene in
the region. Therefore, we performed a linkage analysis on the
352 pedigrees (77% of the whole sample, denoted LODþ)
which showed elevated allele sharing (LOD . 0) over the
region 60–90 cM (covering the LOD-1 region for the
linkage analysis on the whole sample, the location where
therewas maximum evidence
between sites and the maximum LOD score locations from
the samples studied by groups 1–3). The results are shown
in Figure 2. Such an analysis is clearly invalid as a test for
linkage, but may give insights into the likely location of a
disease susceptibility locus. As expected, the magnitude of
the peak (18.0) in the LODþsample far exceeds that in the
whole sample. The shape of the LOD score curve for the
LODþ sample is similar to that from the whole sample
(Fig. 2A) and shows no evidence of there being more than
one disease locus in this region of chromosome 10. Although,
we cannot rule out there being more than one susceptibility
gene within this locus. Likewise, there seems to be no differ-
ence between collection sites in the locations of maximum
IBD allele sharing among the LODþ pedigrees (Fig. 2B),
again supporting the view that the same locus is conferring
risk in all samples. Note that the LODþpedigrees from the
UK sample show the highest IBD, despite the IBD from the
UK sample as a whole being close to 0.5 in the region. This
indicates that a large proportion of pedigrees in this sample
are unlikely to be segregating a disease gene in this region.
The LOD-1 region is reduced to 7 cM (73–80 cM) in the
LODþ sample and this suggests that the LOD scores of
the non-linked pedigrees are randomly distributed across the
Two distinct peaks were observed on chromosome 9. The
maximum LOD of 2.5 was observed at 9q22.33 (104 cM,
nearest marker ¼ D9S910, genome-wide suggestive), and a
second peak with a maximum LOD of 1.2 at 9p21.3 (40 cM,
nearest markers ¼ D9S1870and D9S741).A testforhomogen-
eity of allele sharing was performed between the five sites of
collection. A significant difference in allele sharing was
observed (chromosome-wide P ¼ 0.04). Including the site of
collection as a covariate maximized the LOD at 102 cM.
Inspection of the allele-sharing estimates at this maximum
gave UK: 0.45, NIMH 50: 0.58, NIMH 51: 0.62, NIMH 52:
0.59 and NIA: 0.50, indicating that the linkage evidence at
the main peak is due to the NIMH data only. There was no stat-
istical evidence of a difference between the three original
NIMH collection sites (chromosome-wide P ¼ 0.78). Given
that the test statistic uses the maximum LOD scores on the
chromosome, these significance levels are related to the main
peak on 9q22.33. We also performed supplementary homo-
geneity tests on the restricted chromosome length between
24 and 64 cM, thereby targeting the secondary peak in the
full sample. There was no significant difference observed
between the five collection sites (chromosome-wide P ¼ 0.25).
Analysing the pedigrees with reasonable evidence of
linkage to the main peak (LOD.0 for at least 50% of the
LOD-1 region, peak location +6 cM, n ¼ 208) produced a
single peak at 102 cM. This criterion for linkage evidence
was chosen to ensure definite membership to this peak
rather than the secondary peak. Similarly, focusing on the sec-
ondary peak +6 cM region (n ¼ 199) produced a single peak
at 40 cM. These two peaks from the separate analyses are of
similar height and in the same locations as those when analys-
ing the full data set. Analysing the pedigrees with evidence for
linkage (LOD . 0) for at least 50% of both the main and sec-
ondary peak regions (n ¼ 102) gave LOD scores in both peak
regions, approximately half the height of those seen in the
separate analyses (from approximately half the number of
pedigrees; Fig. 3). These results suggest that pedigrees with
increased allele sharing at one of the loci have a 50%
chance of having increased sharing at the other, i.e. the two
loci are independent. The correlation in allele sharing
between the two loci is not significantly different from what
you would expect by chance, given the distance between the
two loci (P ¼ 0.13), suggesting that the two loci should be
A maximum LOD score of 2.0 was observed at 19q13.32
(70 cM, nearest marker ¼ D19S412, genome-wide sugges-
tive). The LOD-1 region covers 44 cM (46–82 cM), including
Human Molecular Genetics, 2007, Vol. 16, No. 222705
by guest on June 7, 2013
the location of the APOE gene. Similar evidence was observed
in the two original data sets and no evidence of site of collec-
tion heterogeneity was detected. In the ARPs with data, evi-
dence for linkage, conditional on age at onset, produced an
increase in maximum LOD score of 3.8, from 2.7 to 6.3 at
70 cM, close to APOE. Increased allele sharing was observed
in the pairs with a younger age at onset. APOE is known to
have an effect on age at onset (27), so we corrected age at
onset for APOE4 and repeated the analysis. The maximum
age at onset covariate LOD score was reduced, but the
observed effect was not eliminated (Fig. 4). This confirms
that APOE has an effect on age at onset, but also suggests
that there may be an additional locus in the region influencing
age at onset.
ARP locus–locus interaction analyses of chromosomes 9, 10
and 19 were performed, by testing the allele-sharing probabil-
ities of chromosome I, conditional on chromosome J. The
results are presented in Figure 5. The most interesting result
was observed when analysing chromosome 10 (130 cM), con-
ditional on chromosome 19 (38 cM). The unconditional LOD
score at this point on chromosome 10 was 0.2, increasing by
1.8 to 2.0 when conditioning on chromosome 19 (point-wise
P ¼ 0.005). The covariate parameter is positive, indicating
that at these particular locations, ARPs with increased allele
sharing on chromosome 19 also have increased allele
sharing on chromosome 10. A similar effect was observed
when conditioning chromosome 10 on 19. The ARPs with
increased allele sharing in this same region on chromosome
10 also have increased allele sharing on chromosome 9 at
66 cM, although there is no such effect between chromosomes
9 and 19.
A maximum LOD score of 0.2 at 26 cM was observed on
12p13.2 (co-localized by D12S391 and D12S358). Although
we do not observe evidence to suggest that a gene for AD
resides on this chromosome, some studies have observed
such evidence (28,29), and so we performed further covariate
analyses to see whether we could replicate the findings
observed in a sample, including a large proportion of NIMH
pedigrees, of Liang et al. (28). Liang et al. found linkage
Figure 2. Chromosome 10 linkage analyses of pedigrees showing evidence for
linkage (LODþ, i.e. those pedigrees with LOD . 0) between 60 and 90 cM.
(A) Comparison of the peak region. One can see that the peak from the
LODþpedigrees (dashed line) shows a better resolution than that from all
the pedigrees (solid line). (B) IBD allele-sharing probabilities in the LODþ
sample, split by sites of collection.
Figure 3. Chromosome 9 linkage analyses of pedigrees showing evidence for
linkage (LOD.0 for at least 50% of the given region) at solid line, secondary
peak +6 cM (n ¼ 199); dashed line, main peak LOD-1 region (12 cM,
n ¼ 208) and dashed-dotted line: both secondary peak +6 cM and main
peak LOD-1 region (n ¼ 102).
Figure 4. Chromosome 19 linkage analysis of individuals with age at onset
and APOE data. The LOD scores represent an unconditional linkage analysis
(solid line), conditional on age at onset (dashed line), conditional on age at
onset after removing the effect of the presence or absence of APOE4 (dotted-
2706 Human Molecular Genetics, 2007, Vol. 16, No. 22
by guest on June 7, 2013
covariate effects in the 51–67 cM region when conditioning
on the LOD score of an SNP (rs7070570) in the a-catenin
gene (VR22) on chromosome 10 (82.5–86.2 cM, point-wise
P , 0.0001), D9S741 (42 cM, point-wise P ¼ 0.20) and the
proportion of APOE e4 alleles in the pair (point-wise
P ¼ 0.04) in an ordered subset analysis (OSA). We do not
observe significant evidence of these effects in our sample,
observing point-wise significance levels for the three covari-
ates of 0.54, 0.39 and 0.58, respectively. The OSA method
employed by Liang et al. seeks to identify a subset of the
families in the sample defined by a continuous family-wise
covariate that maximizes the evidence for linkage. Our
method uses pair-wise measures of linkage on the chromo-
some being conditioned on, and so might be more sensitive
in the presence of within-family linkage heterogeneity. Never-
theless, if these are true effects, we would also hope to identify
them in our large sample.
Including APOE as a covariate maximized the LOD score at
22 cM (chromosome-wide P ¼ 0.53), the IBD sharing prob-
abilities observed at the peak were 2/2: 0.58, 2/þ: 0.48
and þ/þ: 0.50 in 104, 119 and 374 ARPs, respectively.
Although the APOE effect was not significant, elevated
sharing in the 2/2 pairs was observed, confirming that
found by group 1 (16). The IBD estimate in the APOE 2/2
pairs is of comparable magnitude with that observed at the
peak on chromosome 10, although the sample size is much
The mean (SD) genotyping consistency rate per marker was
estimated to be 97.3 (2.6)%. A total of 46 012 genotypes
were checked over 99 markers. In the four regions identified,
the rate was estimated to be 99.4%, indicating that the geno-
typing in these regions is of a very high quality. Three outliers
were detected with a rate , 90%, namely, D4S403 (26 cM),
D9S1838 (164 cM) and D12S1045 (161 cM), which were
removed from the analysis.
Figure 5. Interaction plots. Increases in LOD score on primary chromosome (x-axis) when allowing for allele sharing on a secondary chromosome (y-axis).
Contour grades increment every increase in LOD of 0.5.
Human Molecular Genetics, 2007, Vol. 16, No. 222707
by guest on June 7, 2013
Multiple genome scans have found evidence for linkage to
chromosomes 9, 10 and 19 (14–21). Interpretation of these
data is difficult because the samples analysed overlapped in
a number of instances, and so the results cannot be considered
to be independent replications. We have amalgamated the raw
data from three large studies of LOAD. Access to the raw data
allowed us to perform a genome screen for linkage to LOAD,
using the same inclusion criteria for all individuals. Further
linkage analyses were also performed in regions of interest
to investigate any potential heterogeneity between sites of col-
lection, locus–locus interactions, age at onset effect and in
particular with APOE. The main aim was to provide the best
estimates for the most likely locations of genes involved
with LOAD. We also aimed to assess the reliability of geno-
typing between the research groups. Although the data ana-
lysed here constitute the largest single sample yet assembled
for a linkage study of AD on chromosome 10, there are
other samples that we did not analyse. These include both
samples of Caucasian origin (20,30) and samples drawn
from other racial groups (21). An analysis of all available
AD scans, allowing for differences in ethnic origin and other
clinical phenotypes, would be worthwhile.
In our large sample, we observe evidence for linkage at
9p21.3, 9q22.33, 10q21.2 and 19q13.32. There is significant
evidence to support a susceptibility locus on 10q21.2, with a
strong indication that this locus operates in a number of popu-
lations, and the most likely location of the risk gene(s) is at
78 cM. Our findings also fail to support the presence of
second locus on 10q. Although these findings are potentially
of considerable importance in guiding further genetic studies
aimed at identifying a susceptibility gene in this region,
some caveats must be noted. In particular, we cannot defini-
tively exclude the possibility that more than one locus for
AD exists on chromosome 10q in spite of the fact that we
did not obtain significant evidence for heterogeneity. The
power of any linkage analysis aimed at identifying a second
(or even third) locus in the peak region, especially without
knowing a definite location of the first, is related to the dis-
tances between loci. Differentiating between one locus or
two in close proximity is difficult in this setting. As the dis-
tance between the two potential loci increases, the chance of
separating their respective linkage signals also increases.
Larger samples of pedigrees genotyped and pooled into a
single linkage analysis (as presented here) will be beneficial,
as will be scans with higher genetic information content and
resolution (e.g. using SNP genotypes). Although we cannot
exclude the possibility that susceptibility is conferred by two
genes located close together, or maybe on either side of the
peak, but outside the LOD-1 region, considerations of parsi-
mony encourage us to suggest that a single locus between
68 and 85 cM is responsible for the linkage to this region.
It has been suggested that the chromosome 10 linkage is
subject to a parent of origin effect, with elevated allele
sharing observed only in a subset pedigrees where AD has
been inherited from the mother’s family (15,31). This
finding is based on a subset of the data analysed in the
present study. However, since data suggesting the parent of
origin of familial risk are only available for two of the
NIMH family sets (sites 50 and 52), we were not able to sup-
plement this analysis. Interestingly, the linkage obtained in
this parent of origin analysis maximized at 75.6 cM, only
2 cM from our peak. Linkage to 10q21.2 (LOD ¼ 1.7) was
observed in a sample containing 457 NIMH pedigrees with
age at onset .50 years (17). The evidence for linkage
appeared to originate from both early- and late-onset cases.
Correspondingly, in agreement with our results, Holmans
et al. (32). found negligible effects of age at onset on the
chromosome 10 linkage in the combined UK/NIMH/NIA
sample of Myers et al. (16) when age at onset was restricted
to be ?65 years. Modest evidence for linkage was observed
to 10q21.2 in an independent sample of predominantly
late-onset (mean age at onset ¼ 70.7, range ¼ 40–89 years)
AD families, using age at onset (rather than AD affection
status) as the trait of interest (30). Linkage to 10q21.2 was
not replicated by Sillen et al. (20) in a small sample of
onset ¼ 69 + 6.8 years), nor by Pericak-Vance et al. (18) in
a sample containing 286 NIMH and 118 NIA pedigrees,
together with 62 pedigrees from CAP (Collaborative Alzhei-
mer Project; Duke, UCLA, Vanderbilt, USA), (family mean
age at onset .60 years, LOD , 1). As we do not have
access to the entire raw data of Pericak-Vance et al. (18), it
is impossible to say with certainty why, in an overlapping
sample, the evidence for linkage to 10q21.2 differed. A com-
bination of many factors could be the cause. The additional
pedigrees used by Pericak-Vance et al. (18) could show no
linkage, thus diluting the signal. The inclusion criteria of
family mean onset of 60 years by Pericak-Vance et al. (18),
rather than individual onset age used here, could reduce the
power to detect linkage to a disease locus that affects those
individuals with later onset. However, the Blacker et al. (17)
analysis that was not constrained by onset age did show evi-
dence for linkage to this region. Another possibility is that
the genetic information content in the Pericak-Vance et al.
(18) sample is less than that in the data analysis presented
here. Lee et al. (21) found evidence of linkage to 10q at
138 cM (NPL ¼ 2.02) with minimal linkage evidence to
10q21.2 (NPL , 1) in a sample of Caribbean-Hispanic pedi-
grees with predominantly LOAD (mean age at onset ¼ 73.5
years). Interestingly, we observe some linkage evidence
nearby (?130 cM) when interaction with chromosome 19 is
taken into account. Thus, although there is good evidence
for linkage to chromosome 10q, there is also likely linkage
heterogeneity due to ethnic origin. Variation in age at onset
and interaction with other loci may also influence linkage
The two linkages on chromosome 9, at 9p21.3 and 9q22.33,
appear to reflect two susceptibility loci that act independently.
Evidence exists for site of collection heterogeneity in the
allele-sharing probabilities. In particular, the main peak on
9q22.33 appears to be contributed primarily by elevated
allele sharing from the NIMH sample. Pericak-Vance et al.
(18) identified a linkage peak close to the one we observed
on 9p22.1 (43 versus 40 cM) in a sample that included 286
NIMH and 118 NIA pedigrees out of their total of 466
pedigrees, with a maximum LOD score of 2.97. In our analy-
sis, linkage evidence to 9p21.3 comes mainly from the NIMH
sites 50 and 51 and the NIA samples (although a formal test of
2708Human Molecular Genetics, 2007, Vol. 16, No. 22
by guest on June 7, 2013
IBD differences across all five samples was non-significant).
Since our sample overlaps with that of Pericak-Vance, this
cannot be considered as evidence of replication. Blacker
et al. report evidence for linkage at 55 cM (15 cM more cen-
tromeric, LOD ¼ 1.3). Wijsman et al. and Sillen et al. found
no linkage evidence to this region on chromosome 9,
whereas Lee et al. report minimal evidence (NPL , 1).
The 9p22.1 region was the second strongest area of linkage
in the Pericak-Vance study, second only to an analysis of the
APOE locus itself. This group also obtained LOD scores in
excess of 1 in the full sample at 4q32.1 (LOD ¼ 1.30),
7q31.31 (LOD ¼ 1.56) and 19q13.3 (LOD ¼ 2.21), although
the evidence decreased when restricting the analyses to
families with at least one autopsy-confirmed AD case. We
do not replicate these findings on 4q32.1 and 7q31.31. The
linkage evidence on 9q22.33 was first identified in the
NIMH sample by Kehoe et al. (14) and subsequently con-
firmed in overlapping late-onset samples (16,17,33). Again,
in agreement with our results, Holmans et al. (32) found no
evidence of an age-at-onset effect on linkage in AD cases
with age at onset ?65years. No evidence for linkage was
reported on 9q22.33 by Pericak-Vance et al., Lee et al. or
Sillen et al. Linkage evidence on chromosome 9 appears to
show inter-sample heterogeneity.
We investigated the possibility that there are other linked
variants in addition to APOE4 that contribute to the LOD
score of 2.0 on 19q13.32, using the software LAMP (34).
The test for other linked variants is not significant, although
the power of this test is low for a LOD of this magnitude
and when APOE is in such close proximity to the peak.
Assuming that APOE4 is the only linked variant in this
region, the population attributable fraction associated with
APOE4 was estimated to be 0.36, with penetrances similar
to an additive model.
The inclusion of APOE genotypes as covariates had no sig-
nificant effects on the linkage analysis in any of the other
linkage regions. This suggests that statistical interactions
between APOE and genes in these regions do not have a
large effect on risk of AD, although it does not preclude the
possibility of biological interactions (35). We also included
IBD sharing probabilities in the regions of interest as covari-
ates in order to search for potential interaction effects on
AD risk between disease susceptibility loci in these regions.
A region on chromosome 10 (?130 cM) was identified by
this analysis when IBD probabilities on chromosome 19
were used as covariates. The region on chromosome 19 impli-
cated in this interaction is at 38 cM from the pter, ?35 cM
from APOE itself. Although ARP analyses do not always
give accurate estimates of disease-locus location (36), it
seems unlikely that the interaction on 19p involves APOE
(on 19q). This conclusion is reinforced by the lack of evidence
for interaction when the APOE genotypes were used as covari-
ates, despite the strong association of APOE with LOAD.
Interestingly, the region on 19p showing interaction evidence
to chromosome 10 is close to that identified by Wijsman et al.
(30). They also concluded that linkage to this region on 19p
did not reflect APOE effects.
We observed excellent genotype reliability between the
three research groups. For diseases with a late age at onset,
genome scans using mainly affected sibling pairs (ASPs) are
common as parental DNAs are often unavailable. Genotyping
error is difficult to detect in small pedigrees (e.g. one ASP)
with no parental genotypes, as all genotypic configurations
are consistent with Mendelian inheritance. The presence of
random genotyping errors is likely to reduce the mean IBD
allele sharing and hence the evidence for linkage. By remov-
ing genotypes that were discrepant between the three samples,
we have produced a cleaner more powerful sample for
Despite the strength of the linkage finding on chromosome
10, it has not been possible to identify a susceptibility locus.
Indeed, with the exception of APOE, no such risk variants
have been identified for LOAD. There are number of possible
reasons for this. First, a number of interacting susceptibility
genes could reside within a linked region, of which the mar-
ginal effects are small and hence difficult to detect through tra-
ditional single marker tests of association. Second, AD shows
a degree of clinical heterogeneity (37) which might relate to
the underlying genetic architecture. It follows that difficulties
in replicating genetic findings might result from sample differ-
ences and the field might benefit from more careful attention
to phenotype definition. Third, evidence from the linkage
analysis of Pericak-Vance et al. (18) also suggests the possi-
bility of differences in allele-sharing estimates between clini-
cal and neuropathological diagnoses of AD. Extended
phenotypic data are not always available and sometimes
inconsistently reported. Further analyses including these data
will therefore only be possible following extensive data
quality assessment, a subject of future work.
To conclude, our analyses show evidence for disease loci
for LOAD on chromosome 10q (near 78 cM), chromosome
9q (near 104 cM), chromosome 19q (near 70 cM; likely to
reflect the APOE gene) and, less convincingly, chromosome 9p
(near 40 cM). However, there is evidence for linkage hetero-
geneity between samples, both directly, from our own ana-
lyses, and possibly indirectly, from the failures of other
studies to replicate the linkages on chromosomes 9 and 10.
This heterogeneity may also be influenced by other factors
such as variation in age at onset and may explain the moderate
evidence for association so far found to genes in these regions.
Thus, there is a need for a large-scale analysis using all avail-
able LOAD linkage samples in order to ascertain which
regions are most likely to contain AD susceptibility genes.
These regions should then be subjected to detailed analysis
in large association studies, using clinical covariates such as
age at onset, to reduce the effects of heterogeneity.
MATERIALS AND METHODS
The data analysed were generated by three research groups.
Group 1 genotyped pedigrees from the UK (n ¼ 82), NIMH
(n ¼ 212) and NIA (n ¼ 72) (14,16,25,32). Group 2 genotyped
pedigrees from the NIMH (n ¼ 437) (17,24). Group 3 also
genotyped pedigrees from the NIMH (n ¼ 287) with geno-
types limited to chromosome 10 only (15,31). The three data
sets have the NIMH sample in common, with 196 (64%) pedi-
grees and 516 (21%) individuals in this overall NIMH sample
genotyped by all three groups. It should be noted that although
Human Molecular Genetics, 2007, Vol. 16, No. 222709
by guest on June 7, 2013
all three groups had access to the NIMH database of individ-
uals, groups 2 and 3 had access to additional pedigrees that are
not available on the NIMH database. In total, three collection
sites contributed data to the NIMH database, and they are sites
50 (University of Alabama- Birmingham, AL, USA), 51 (Mas-
sachusetts General Hospital, MA, USA) and 52 (John
Hopkins, MD, USA), all based in the USA. A summary of
the number of pedigrees and ASPs from each sample is pre-
sented in Table 2. In addition to the ASPs, 57 non-sibling
ARPs informative for linkage were also included in the analy-
sis. To minimize heterogeneity, the samples were limited to
individuals of Caucasian origin. The original analyses by
groups 2 and 3 included pedigrees of non-Caucasian ancestry—
these were excluded from the current analysis.
Individuals were considered affected if they were diagnosed
with definite or probable AD using the NINCDS-ADRDA
(38) criteria, with an age at onset of at least 60 years. Previous
presentations of the late-onset data have used a cut-off of 65
years. We relaxed this criterion to increase power, as we do
not believe that this will include a significant proportion of
familial autosomal dominant cases. Indeed, there was only
one ARP with multiple reports of early onset AD cases
(onset , 65 years) in the same family. The original report
by group 2 also included an analysis with onset ages of at
least 50 years and an early/mixed onset analysis, and group 3
only presented analyses including cases with the full onset
age range. The two groups also included some individuals
with possible rather than probable or definite AD—these
were also excluded from the current analysis.
Group 1 first genotyped 237 microsatellite markers (average
spacing ¼ 20 cM)
maximum LOD score .1 were followed up with a further
91 markers (peak region average marker spacing ¼ 10 cM)
(16,25). Group 2 genotyped 381 microsatellite markers
(average spacing ¼ 9 cM). Group 3 genotyped a further 18
markers on chromosome 10 in the families ascertained at
NIMH sites 50 and 52 (0–80 cM), plus an additional seven
markers in the families from site 52 (0–54 cM). When the
20 markers genotyped in the NIMH families by groups 1
and 2 were taken into account, it gave a total of 45 markers
on chromosome 10 (average spacing ¼ 4 cM). Detailed infor-
mation on the laboratory methods used to establish the micro-
satellite genotypes by each group are provided in their
respective original publications (14,16,17,31). The genotypes
from each group were provided in standard pedfile format,
one file per chromosome. For each chromosome, one large
pedfile was created by joining the original pedfiles for the
given chromosome from each group, based on a standardized
marker map comprising all markers. This map was derived
from Marshfield (Marshfield Clinic, Marshfield genetic map,
supplemented with information from Decode (39) and the
UCSC Human Genome Browser (www.genome.ucsc.edu,
May 2004 assembly, build 35) where necessary. Where the
(14). The16 peak regions witha
same genetic marker was genotyped by more than one
group, each of the original markers remained as separate
markers in the final pedfile, separated by an arbitrary distance
of 0.001 cM. Individuals not genotyped at specific markers
were coded as missing data, i.e. given a 0–0 genotype. This
method of sample amalgamation allowed analysis of the
complete data set and made no assumption that the allele
numbering was consistent between groups.
Method of analysis
Given the availability of genotypic data, both within and
between family relationships were confirmed statistically;
see original publications for further information. Sample
specific marker allele frequencies were estimated with
SPLINK (40). Given that the data from all three groups com-
prised individuals from the NIMH sample, when the same
genetic markers were genotyped by more than one group,
we were able to check the reliability of the genotyping. As
the genotypes from groups 2 and 3 were the same, quality
control checks were performed to compare genotypes from
group 1 with the combination of groups 2 and 3. Discrepant
genotypes were removed from the analysis. One pedigree
(53 bits) comprising three ASPs was trimmed to permit analy-
sis within the constraints of modern computing power. Multi-
point model-free ARP linkage analysis was performed to give
likelihood ratio LOD scores for all chromosomes under the
logistic regression model suggested by Rice (41), using IBD
allele-sharing probabilities estimated at 2 cM intervals with
MERLIN (42,43). Care was taken to consider marker–
marker LD, most of which was imposed when the three
overlapping samples were combined, i.e. the inclusion of
duplicated markers, which were set to be 0.001 cM apart.
Markers separated by ,0.5 cM were grouped into clusters
and estimated haplotype frequencies used in the IBD esti-
mation to assume linkage disequilibrium within each cluster.
For improved accuracy in regions of interest, the allele-sharing
probabilities were estimated at 1 cM intervals. The levels of
significant and suggestive evidence for linkage employed
were as defined by Lander and Kruglyak (26).
Furthertests were performedon chromosomeswith evidence
of linkage (LOD . 1). Separate linkage analyses of the data
from the three groups were performed. The results may differ
from previous published analyses as the method of analysis
and selection criteria of affected individuals differ. In addition,
data quality checks removed some questionable and hence
tion that the parental alleles were inherited independently, the
Table 2. The total number of ASPs (pedigrees) for each sample, split by site of
Sample NIMH 50NIMH 51NIMH 52UKNIA Total
109 (82) 90 (72)486 (366)
109 (82) 90 (72)
109 (82)90 (72)
2710 Human Molecular Genetics, 2007, Vol. 16, No. 22
by guest on June 7, 2013
probability of allele sharing IBD was modelled using logistic
regression (41,44,45). Site of collection (i.e. UK, NIMH 50,
For each 2 cM position, a LOD score was obtained for the
models (i) with and (ii) without the site of collection variables.
Linkage peaks from standard analyses are often some distance
from the disease loci (36); therefore, to allow for the multiple
testing of more than one 2 cM position on the chromosome,
the maximum LOD score for models (i) and (ii) were obtained
over all positions analysed on the chromosome. The difference
A value of Tobs¼ 0 implies no difference in models (i) and (ii),
i.e. there is no site of collection effect. A value of Tobs. 0
suggests some site of collection effect, the significance of
which was assessed through simulations. The original site of
collection data were permuted over all pedigrees and a value
of Tsimfor the simulated data set produced. This was repeated
n ¼ 10 000times.Anestimateofthechromosome-widesignifi-
canceofthetestforhomogeneitywasassessedwithP ¼ (r þ 1)/
(n þ 1), where r is the number of times Tsimexceeded Tobs.
A similar test was performed in order to condition linkage by
APOE genotype. Each affected individual was scored as 2 or
þ, indicating whether they carried at least one copy of the
APOE4 allele (þ) or not (2). Pairs of affected individuals
were then constructed which could be either 2/2, 2/þ or
þ/þ, where a 2/2 (þ/þ) pair indicates that neither (both)
members of the pair carried at least one APOE4 allele,
whereas a 2/þ pair indicates that just one member had an
APOE4 allele. The APOE covariate was entered into the
regression model as two indicator variables. Constraints were
applied to ensure that the allele sharing of the discordant
pairs could not exceed that of the concordant pairs. This test
used the same method as that presented in Myers et al. (16).
As APOE resides on 19q, no APOE covariate test was per-
measures, such as age at onset and allele-sharing probabilities,
into an analysis is also possible. Conditioning on allele-sharing
probabilities allows us to investigate the possibility of inter-
actions between two chromosomes. This method is considered
to be less powerful than conditioning on the actual risk factor
itself (e.g. the risk allele) (46), although until these risk
factors are identified, conditioning on allele-sharing probabil-
ities is considered to be a useful substitute. A test of the allele-
sharing probabilities on chromosome I, conditional on J was
performed. For completeness, a test for chromosome J, con-
ditional on I was also performed, which gave similar results.
Supplementary Material is available at HMG Online.
Conflict of Interest statement. None.
This analysis was supported primarily by the Medical
Research Council (UK) who support the Medical Research
Council Co-operative group in Neuropsychiatric Genetics
(M.L.H., M.J.O., J.W., M.O.D.), and a programme of AD
research (J.W., M.J.O., M.O.D., P.H., S.L.). It was also
supported by the Alzheimer’s Research Trust (M.J.O., J.W.).
The original data collection was supported by the Medical
Research Council (UK) and by the National Institute on
Aging, National Cell Repository for Alzheimer’s Disease
and National Institute of Mental Health Genetics Initiative
for Alzheimer’s Disease.
1. Gatz, M., Reynolds, C.A., Fratiglioni, L., Johansson, B., Mortimer, J.A.,
Berg, S., Fiske, A. and Pedersen, N.L. (2006) Role of genes and
environments for explaining Alzheimer disease. Arch. Gen. Psychiatry,
2. Nussbaum, R.L. and Ellis, C.E. (2003) Alzheimer’s disease and
Parkinson’s disease. N. Engl. J. Med., 348, 1356–1364.
3. Fratiglioni, L., De Ronchi, D. and Aguero-Torres, H. (1999) Worldwide
prevalence and incidence of dementia. Drugs Aging, 15, 365–375.
4. Goate, A., Chartier-Harlin, M.C., Mullan, M., Brown, J., Crawford, F.,
Fidani, L., Giuffra, L., Haynes, A., Irving, N., James, L. et al. (1991)
Segregation of a missense mutation in the amyloid precursor protein gene
with familial Alzheimer’s disease. Nature, 349, 704–706.
5. Levy-Lahad, E., Wasco, W., Poorkaj, P., Romano, D.M., Oshima, J.,
Pettingell, W.H., Yu, C.E., Jondro, P.D., Schmidt, S.D., Wang, K. et al.
(1995) Candidate gene for the chromosome 1 familial Alzheimer’s disease
locus. Science, 269, 973–977.
6. Sherrington, R., Rogaev, E.I., Liang, Y., Rogaeva, E.A., Levesque, G.,
Ikeda, M., Chi, H., Lin, C., Li, G., Holman, K. et al. (1995) Cloning of a
gene bearing missense mutations in early-onset familial Alzheimer’s
disease. Nature, 375, 754–760.
7. Pedersen, N.L., Gatz, M., Berg, S. and Johansson, B. (2004) How
heritable is Alzheimer’s disease late in life? Findings from Swedish twins.
Ann. Neurol., 55, 180–185.
8. Bales, K.R., Dodart, J.C., DeMattos, R.B., Holtzman, D.M. and Paul, S.M.
(2002) Apolipoprotein E, amyloid, and Alzheimer disease. Mol. Interv., 2,
9. Henderson, A.S., Easteal, S., Jorm, A.F., Mackinnon, A.J., Korten, A.E.,
Christensen, H., Croft, L. and Jacomb, P.A. (1995) Apolipoprotein E
allele epsilon 4, dementia, and cognitive decline in a population sample.
Lancet, 346, 1387–1390.
10. Martins, R.N., Clarnette, R., Fisher, C., Broe, G.A., Brooks, W.S.,
Montgomery, P. and Gandy, S.E. (1995) ApoE genotypes in Australia:
roles in early and late onset Alzheimer’s disease and Down’s syndrome.
Neuroreport, 6, 1513–1516.
11. Kukull, W.A., Schellenberg, G.D., Bowen, J.D., McCormick, W.C.,
Yu, C.E., Teri, L., Thompson, J.D., O’Meara, E.S. and Larson, E.B.
(1996) Apolipoprotein E in Alzheimer’s disease risk and case detection: a
case-control study. J. Clin. Epidemiol., 49, 1143–1148.
12. Daw, E.W., Payami, H., Nemens, E.J., Nochlin, D., Bird, T.D.,
Schellenberg, G.D. and Wijsman, E.M. (2000) The number of trait loci
in late-onset Alzheimer disease. Am. J. Hum. Genet., 66, 196–204.
13. Bertram, L., McQueen, M.B., Mullin, K., Blacker, D. and Tanzi, R.E.
(2007) Systematic meta-analyses of Alzheimer disease genetic association
studies: the AlzGene database. Nat. Genet., 39, 17–23.
14. Kehoe, P., Wavrant-De Vrieze, F., Crook, R., Wu, W.S., Holmans, P.,
Fenton, I., Spurlock, G., Norton, N., Williams, H., Williams, N. et al.
(1999) A full genome scan for late onset Alzheimer’s disease. Hum. Mol.
Genet., 8, 237–245.
15. Bassett, S.S., Avramopoulos, D., Perry, R.T., Wiener, H., Watson, B., Jr,
Go, R.C. and Fallin, M.D. (2006) Further evidence of a maternal
parent-of-origin effect on chromosome 10 in late-onset Alzheimer’s
disease. Am. J. Med. Genet. B Neuropsychiatr. Genet., 141B, 537–540.
16. Myers, A., Wavrant De-Vrieze, F., Holmans, P., Hamshere, M., Crook, R.,
Compton, D., Marshall, H., Meyer, D., Shears, S., Booth, J. et al. (2002)
Full genome screen for Alzheimer disease: stage II analysis. Am. J. Med.
Genet., 114, 235–244.
17. Blacker, D., Bertram, L., Saunders, A.J., Moscarillo, T.J., Albert, M.S.,
Wiener, H., Perry, R.T., Collins, J.S., Harrell, L.E., Go, R.C. et al. (2003)
Results of a high-resolution genome screen of 437 Alzheimer’s disease
families. Hum. Mol. Genet., 12, 23–32.
Human Molecular Genetics, 2007, Vol. 16, No. 222711
by guest on June 7, 2013
18. Pericak-Vance, M.A., Grubber, J., Bailey, L.R., Hedges, D., West, S., Download full-text
Santoro, L., Kemmerer, B., Hall, J.L., Saunders, A.M., Roses, A.D. et al.
(2000) Identification of novel genes in late-onset Alzheimer’s disease.
Exp. Gerontol., 35, 1343–1352.
19. Rademakers, R., Cruts, M., Sleegers, K., Dermaut, B., Theuns, J.,
Aulchenko, Y., Weckx, S., De Pooter, T., Van den Broeck, M., Corsmit, E.
et al. (2005) Linkage and association studies identify a novel locus for
Alzheimer disease at 7q36 in a Dutch population-based sample.
Am. J. Hum. Genet., 77, 643–652.
20. Sillen, A., Forsell, C., Lilius, L., Axelman, K., Bjork, B.F., Onkamo, P.,
Kere, J., Winblad, B. and Graff, C. (2006) Genome scan on Swedish
Alzheimer’s disease families. Mol. Psychiatry, 11, 182–186.
21. Lee, J.H., Mayeux, R., Mayo, D., Mo, J., Santana, V., Williamson, J.,
Flaquer, A., Ciappa, A., Rondon, H., Estevez, P. et al. (2004) Fine
mapping of 10q and 18q for familial Alzheimer’s disease in Caribbean
Hispanics. Mol. Psychiatry, 9, 1042–1051.
22. Bertram, L., Blacker, D., Mullin, K., Keeney, D., Jones, J., Basu, S., Yhu,
S., McInnis, M.G., Go, R.C., Vekrellis, K. et al. (2000) Evidence for
genetic linkage of Alzheimer’s disease to chromosome 10q. Science, 290,
23. Ertekin-Taner, N., Graff-Radford, N., Younkin, L.H., Eckman, C., Baker,
M., Adamson, J., Ronald, J., Blangero, J., Hutton, M. and Younkin, S.G.
(2000) Linkage of plasma Abeta42 to a quantitative locus on chromosome
10 in late-onset Alzheimer’s disease pedigrees. Science, 290, 2303–2304.
24. Li, Y.J., Scott, W.K., Hedges, D.J., Zhang, F., Gaskell, P.C., Nance, M.A.,
Watts, R.L., Hubble, J.P., Koller, W.C., Pahwa, R. et al. (2002) Age at
onset in two common neurodegenerative diseases is genetically
controlled. Am. J. Hum. Genet., 70, 985–993.
25. Myers, A., Holmans, P., Marshall, H., Kwon, J., Meyer, D., Ramic, D.,
Shears, S., Booth, J., DeVrieze, F.W., Crook, R. et al. (2000)
Susceptibility locus for Alzheimer’s disease on chromosome 10. Science,
26. Lander, E. and Kruglyak, L. (1995) Genetic dissection of complex traits:
guidelines for interpreting and reporting linkage results. Nat. Genet., 11,
27. Farrer, L.A., Cupples, L.A., Haines, J.L., Hyman, B., Kukull, W.A.,
Mayeux, R., Myers, R.H., Pericak-Vance, M.A., Risch, N. and van Duijn,
C.M. (1997) Effects of age, sex, and ethnicity on the association between
apolipoprotein E genotype and Alzheimer disease. A meta-analysis.
APOE and Alzheimer Disease Meta Analysis Consortium. JAMA, 278,
28. Liang, X., Schnetz-Boutaud, N., Kenealy, S.J., Jiang, L., Bartlett, J.,
Lynch, B., Gaskell, P.C., Gwirtsman, H., McFarland, L., Bembe, M.L.
et al. (2006) Covariate analysis of late-onset Alzheimer disease refines the
chromosome 12 locus. Mol. Psychiatry, 11, 280–285.
Ciappa, A., Rondon, H.Z., Estevez, P., Lantigua, R. et al. (2002)
Chromosome-12 mapping of late-onset Alzheimer disease among
Caribbean Hispanics. Am. J. Hum. Genet., 70, 237–243.
30. Wijsman, E.M., Daw, E.W., Yu, C.E., Payami, H., Steinbart, E.J.,
Nochlin, D., Conlon, E.M., Bird, T.D. and Schellenberg, G.D. (2004)
Evidence for a novel late-onset Alzheimer disease locus on chromosome
19p13.2. Am. J. Hum. Genet., 75, 398–409.
31. Bassett, S.S., Avramopoulos, D. and Fallin, D. (2002) Evidence for parent
of origin effect in late-onset Alzheimer disease. Am. J. Med. Genet., 114,
32. Holmans, P., Hamshere, M., Hollingworth, P., Rice, F., Tunstall, N.,
Jones, S., Moore, P., Wavrant DeVrieze, F., Myers, A., Crook, R. et al.
(2005) Genome screen for loci influencing age at onset and rate of decline
in late onset Alzheimer’s disease. Am. J. Med. Genet. B Neuropsychiatr.
Genet., 135, 24–32.
33. Perry, R.T., Wiener, H., Harrell, L.E., Blacker, D., Tanzi, R.E., Bertram,
L., Bassett, S.S. and Go, R.C. (2007) Follow-up mapping supports the
evidence for linkage in the candidate region at 9q22 in the NIMH
Alzheimer’s disease Genetics Initiative cohort. Am. J. Med. Genet. B
Neuropsychiatr. Genet., 144, 220–227.
34. Li, M., Boehnke, M. and Abecasis, G.R. (2005) Joint modeling of linkage
and association: identifying SNPs responsible for a linkage signal.
Am. J. Hum. Genet., 76, 934–949.
35. Cordell, H.J. (2002) Epistasis: what it means, what it doesn’t mean, and
statistical methods to detect it in humans. Hum. Mol. Genet., 11,
36. Cordell, H.J. (2001) Sample size requirements to control for stochastic
variation in magnitude and location of allele-sharing linkage statistics in
affected sibling pairs. Ann. Hum. Genet., 65, 491–502.
37. Hollingworth, P., Hamshere, M.L., Moskvina, V., Dowzell, K., Moore,
P.J., Foy, C., Archer, N., Lynch, A., Lovestone, S., Brayne, C. et al.
(2006) Four components describe behavioral symptoms in 1,120
individuals with late-onset Alzheimer’s disease. J. Am. Geriatr. Soc., 54,
38. McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D. and
Stadlan, E.M. (1984) Clinical diagnosis of Alzheimer’s disease: report of
the NINCDS-ADRDA Work Group under the auspices of Department of
Health and Human Services Task Force on Alzheimer’s disease.
Neurology, 34, 939–944.
39. Kong, A., Gudbjartsson, D.F., Sainz, J., Jonsdottir, G.M., Gudjonsson,
S.A., Richardsson, B., Sigurdardottir, S., Barnard, J., Hallbeck, B.,
Masson, G. et al. (2002) A high-resolution recombination map of the
human genome. Nat. Genet., 31, 241–247.
40. Holmans, P. and Clayton, D. (1995) Efficiency of typing unaffected
relatives in an affected-sib-pair linkage study with single-locus and
multiple tightly linked markers. Am. J. Hum. Genet., 57, 1221–1232.
41. Rice, J.P. (1997) The role of meta-analysis in linkage studies of complex
traits. Am. J. Med. Genet., 74, 112–114.
42. Abecasis, G.R., Cherny, S.S., Cookson, W.O. and Cardon, L.R. (2002)
Merlin—rapid analysis of dense genetic maps using sparse gene flow
trees. Nat. Genet., 30, 97–101.
43. Abecasis, G.R. and Wigginton, J.E. (2005) Handling marker–marker
linkage disequilibrium: pedigree analysis with clustered markers.
Am. J. Hum. Genet., 77, 754–767.
44. Rice, J.P. (2001) Diagnosis as a covariate in sib-pair linkage analysis.
Am. J. Med. Genet., 105, 55–56.
45. Hamshere, M.L., Macgregor, S., Moskvina, V., Nikolov, I.N. and
Holmans, P.A. (2005) Covariate linkage analysis of GAW14 simulated
data incorporating subclinical phenotype, sex, population,
parent-of-origin, and interaction. BMC Genet., 6 (Suppl 1), S45.
46. Holmans, P. (2002) Detecting gene-gene interactions using affected sib
pair analysis with covariates. Hum. Hered., 53, 92–102.
2712Human Molecular Genetics, 2007, Vol. 16, No. 22
by guest on June 7, 2013