Genetic analysis of the maximum drinks phenotype

Abstract

Using data provided by the Collaborative Study on the Genetics of Alcoholism we studied the genetics of a quantitative trait: the maximum number of drinks consumed in a 24-hour period. A two-stage method was used. First, linkage analysis was performed, followed by association analysis in regions where linkage was detected. Additionally, the extent of linkage disequilibrium among single-nucleotide polymorphisms (SNP) associated with the phenotype was assessed. Linkage to chromosomes 2 and 7 was detected, and follow-up association analysis found multiple trait-associated SNPs in the chromosome 7 linkage region. Chromosome 4, which has been implicated in previous studies of the maximum drinks phenotype, did not pass our threshold for linkage evidence in stage 1, but secondary analyses of this chromosome indicated modest evidence for both linkage and association. The evidence suggests that chromosome 7 may harbor an additional locus influencing the maximum drinks consumption phenotype.

Full text

Page 1

BioMed Central
Page 1 of 5
(page number not for citation purposes)
BMC Genetics
Open Access
Proceedings
Genetic analysis of the maximum drinks phenotype
Scott F Saccone*1, Nancy L Saccone2, Rosalind J Neuman1,2 and
John P Rice1,2
Address: 1Department of Psychiatry, Washington University School of Medicine, St. Louis, Missouri 63110 USA and 2Department of Genetics,
Washington University School of Medicine, St. Louis, Missouri 63110 USA
Email: Scott F Saccone* - ssaccone@link.wustl.edu; Nancy L Saccone - nsaccone@vodka.wustl.edu; Rosalind J Neuman - roz@gretta.wustl.edu;
John P Rice - john@zork.wustl.edu
* Corresponding author
Abstract
Using data provided by the Collaborative Study on the Genetics of Alcoholism we studied the
genetics of a quantitative trait: the maximum number of drinks consumed in a 24-hour period. A
two-stage method was used. First, linkage analysis was performed, followed by association analysis
in regions where linkage was detected. Additionally, the extent of linkage disequilibrium among
single-nucleotide polymorphisms (SNP) associated with the phenotype was assessed. Linkage to
chromosomes 2 and 7 was detected, and follow-up association analysis found multiple trait-
associated SNPs in the chromosome 7 linkage region. Chromosome 4, which has been implicated
in previous studies of the maximum drinks phenotype, did not pass our threshold for linkage
evidence in stage 1, but secondary analyses of this chromosome indicated modest evidence for both
linkage and association. The evidence suggests that chromosome 7 may harbor an additional locus
influencing the maximum drinks consumption phenotype.
Background
The data provided by the Collaborative Study on the
Genetics of Alcoholism (COGA) for Genetic Analysis
Workshop 14 (GAW14) includes the "maximum number
of drinks consumed in a 24-hour period." This phenotype
is closely related to alcoholism diagnosis, and a previous
genome screen of this phenotype in COGA resulted in evi-
dence for linkage to chromosome 4 in the vicinity of the
alcohol dehydrogenase (ADH) gene cluster [1]. The
GAW14 dataset provides nearly 16,000 genotyped single-
nucleotide polymorphisms (SNP) that were not available
in the original COGA data. For this report we have ana-
lyzed these additional genotypic data to see whether the
original linkage findings can be confirmed or extended.
Linkage signals were followed up with association analy-
sis to make use of the density of the SNP data to refine
linkage signals and potentially localize genetic variants
that may influence alcohol consumption.
Methods
In defining the quantitative trait, we set the phenotype of
individuals who report a maximum drinks value of zero to
be unknown, since these unexposed individuals have
undetermined response to alcohol. We reduced skewness
by taking the logarithm of the maximum number of
drinks. A linear regression was used to correct for sex, and
the final trait value was defined as the residual from the
regression.
Our primary analysis consisted of two stages, applied to
the cleaned SNP data. In the first stage we tested for
genetic linkage using the SNP marker data provided by
Illumina. Nonparametric two-point quantitative trait
from Genetic Analysis Workshop 14: Microsatellite and single-nucleotide polymorphism
Noordwijkerhout, The Netherlands, 7-10 September 2004
Published: 30 December 2005
BMC Genetics 2005, 6(Suppl 1):S124doi:10.1186/1471-2156-6-S1-S124
<supplement> <title> <p>Genetic Analysis Workshop 14: Microsatellite and single-nucleotide polymorphism</p> </title> <editor>Joan E Bailey-Wilson, Laura Almasy, Mariza de Andrade, Julia Bailey, Heike Bickeböller, Heather J Cordell, E Warwick Daw, Lynn Goldin, Ellen L Goode, Courtney Gray-McGuire, Wayne Hening, Gail Jarvik, Brion S Maher, Nancy Mendell, Andrew D Paterson, John Rice, Glen Satten, Brian Suarez, Veronica Vieland, Marsha Wilcox, Heping Zhang, Andreas Ziegler and Jean W MacCluer</editor> <note>Proceedings</note> </supplement>

Page 2

BMC Genetics 2005, 6:S124
Page 2 of 5
(page number not for citation purposes)
linkage analysis was performed using the "--qtl" option of
the program MERLIN [1]. The method compares allele
sharing among individuals with similar phenotypes; indi-
viduals at extreme ends of the distribution are given
greater weight. We chose two-point rather than
multipoint analysis because strong linkage disequilibrium
(LD) among the densely spaced SNP markers could lead
to inflated evidence for linkage if multipoint methods are
used without accounting for marker-marker LD. Markers
exhibiting linkage with a LOD score greater than 1.8 were
used to define regions for further study. In the second
stage a 20-cM interval centered at the marker was used to
test for trait association. SNPs in this interval from the
combined Illumina and Affymetrix sets were analyzed for
trait association using the quantitative pedigree disequi-
librium test (QPDT) [2] as implemented in the program
UNPHASED [3]. Our choice to use the Illumina SNP set
(4,710 markers) for the first stage and the Affymetrix SNP
set (11,120 markers) together with the Illumina set for the
follow-up association analysis allowed us to carry out the
initial genome screen and then proceed with fine-map-
ping in regions of linkage using the more densely spaced
SNPs provided by Affymetrix.
To supplement our primary two-stage analysis, we carried
out some related additional analyses. These are described
below.
For comparison with the stage 1 SNP linkage screen, link-
age analysis of the autosomal microsatellite data was per-
formed using the same two-point nonparametric analysis
as above. We also repeated this microsatellite analysis
using a 2-cM multipoint grid because LD among these
markers was not a significant concern.
For chromosome 4, prior linkage evidence for the maxi-
mum drinks phenotype has been reported [1], making
this a chromosome of special interest. Using the Illumina
markers for chromosome 4, we then tested for linkage in
the first stage using a regression of the estimated IBD sta-
tus against the squared sum and squared difference of the
trait values, as implemented by the program MERLIN-
REGRESS [4]. As this analysis method is computationally
intensive, we applied it only to chromosome 4, allowing
a more comprehensive study of this chromosome of inter-
est.
LD between markers should be considered when using
SNP data to search for disease or trait association. Thus,
SNPs exhibiting trait association in our primary analysis
that were also in close proximity to one another were
tested to see if they were in LD with each other. Pair-wise
LD coefficients were computed using the program
LDMAX, which is part of the GOLD computer package [5];
results are reported using the normalized coefficient D' =
D/|D|max, where D = p11 - p1 p2, p11 is the frequency for the
haplotype containing allele 1 at both markers, pi is the fre-
quency of allele 1 at the ith marker of the pair, and |D|max
is the maximum possible (absolute) value of D given the
marker allele frequencies.
Results
Out of a sample of 1,614 individuals, the maximum
number of drinks was recorded for 1,388 (86%) individ-
uals. The mean, median, and mode were 17.8, 12, and 24,
respectively, with a standard deviation of 17.3. Due to
computer memory constraints, 9 out of 143 families
(6%), consisting of 243/1,614 (15%) individuals, were
automatically dropped by MERLIN from the linkage anal-
ysis in stage 1, and this reduction in sample is a limitation
of our analyses. The average maximum drinks value in the
dropped families was slightly lower than the overall aver-
age (by 0.16 of a standard deviation); this difference does
not appear to have a systematic cause and is not expected
to affect the interpretability of the results.
We used 4,710 markers from the Illumina SNP set. The
mean spacing between markers was 0.77 cM on the
genetic map and 621 kb on the physical map, and the
mean minor allele frequency (MAF) was 0.39. We used
11,120 markers from the Affymetrix SNP set. The mean
spacing was 0.32 cM (258 kb on the physical map), and
the mean MAF was 0.27.
In our primary analysis, only chromosomes 2 and 7 were
found to have LOD scores over 1.8 at stage 1. Tables 1 and
2 show the results of the combined two-point linkage and
trait association analysis (QPDT) for these two chromo-
somes. For the stage 2 association analyses, the tables
report only those SNPs significant at the 0.01 level. We
note that while multiple SNPs throughout the linked
regions were significant at the 0.05 level, only one SNP on
chromosome 2 and multiple SNPs on chromosome 7
Table 1: Two-point linkage (TPL-stage 1) and trait association (QPDT-stage 2) results for chromosome 2
MarkerSourceMAFMbp cMMethod LOD
χ2
p-value
rs1025053
tsc0875114
rs1525351
Illum
Affy
Illum
0.47
0.21
0.44
8.238
8.345
216.466
16.1
21.7
212.9
TPL
QPDT
TPL
1.97
-a
2.68
--
9.05
-
0.003
-
a-, not applicable.

Page 3

BMC Genetics 2005, 6:S124
Page 3 of 5
(page number not for citation purposes)
gave association results significant at the 0.01 level. Table
3 lists the results for the additional two-point regression
analysis performed on chromosome 4 by MERLIN-
REGRESS and the QPDT results computed by
UNPHASED for SNPs in the linkage regions detected by
the regression. The maximal LOD of 2.14 occurred at 97.4
Mb (100.4 cM).
We computed values of D' for the significantly associated
SNPs on chromosomes 2 and 7 (Tables 1 and 2). Thus, we
focused on chromosome 7 because multiple significant
SNPs were detected there. An LD block extended from
124.757 Mb to 124.791 Mb, covering the three most sig-
nificantly associated SNPs; in fact, this block satisfied |D'|
≥ 0.9 for all marker pairs. Thus, the evidence for associa-
tion in this region does not come from three independent
SNPs but rather indicates a single LD block associated
with the phenotype.
Our analyses of the microsatellite data found no evidence
for linkage satisfying our LOD threshold of 1.8 in either
the two-point or the multipoint screens. For the two-point
screen, the only LOD scores over 1.0 occurred on chromo-
some 7 at D7S821 (LOD = 1.13, position = 116.6 cM) and
chromosome 15 at D15S230 (LOD = 1.43, position =
135.6 cM) and D15S642 (LOD = 1.12, position = 152.1
cM). For the multipoint analysis, the only LOD score over
1.0 was 1.49, which occurred on chromosome 15 at 142
cM. Thus, in a direct comparison of the two-point screens
using SNPs versus microsatellites, it is curious that higher
LOD scores were detected with the SNPs than with the
microsatellites, even though the SNPs are individually less
informative. It may be the density of the SNP map that
was beneficial and allowed the detection of linkage evi-
dence that was missed by the microsatellite screen. To
examine this hypothesis, we compared the locations of
the trait-linked SNP markers on chromosomes 2 and 7 to
the locations of the microsatellite markers using inte-
grated map information from the National Center for Bio-
technology Information (NCBI) [6], specifically, the
NCBI build 35.1 reference assembly. This assembly places
the peak SNP on chromosome 7, rs322812, at 127.338
Mb. This falls near the midpoint of the flanking microsat-
ellites, which are located at 122.205 Mb and 131.736 Mb
of the same build, indicating a modest gap, which is also
borne out by the genetic positions of these microsatellites
at 146.7 cM and 156.2 cM on the GAW14 map. However,
for some of the other trait-linked SNPs, there do appear to
Table 2: Two-point linkage (stage 1) and trait association (QPDT-stage 2) results for chromosome 7
Marker SourceMAFMbp cM MethodLOD
χ2
p-value
tsc0229102
rs727708
rs719530
rs846427
rs1860487
tsc0064426
tsc1086900
tsc0058943
rs322812
Affy
Illum
Illum
Illum
Illum
Affy
Affy
Affy
Illum
0.46
0.49
0.42
0.38
0.47
0.37
0.37
0.42
0.44
89.250
102.891
109.451
119.048
123.270
124.757
124.758
124.791
127.299
101.9
114.2
121.8
127.3
129.3
128.6
128.6
128.6
131.3
QPDT
TPL
TPL
TPL
TPL
QPDT
QPDT
QPDT
TPL
-a
7.29
-
-
-
-
7.53
8.23
7.91
-
0.007
-
-
-
-
0.006
0.004
0.005
-
2.28
2.14
2.17
1.91
-
-
-
2.77
a-, not applicable.
Table 3: Two-point regression (stage 1) and trait association (QPDT-stage 2) results for chromosome 4
MarkerSource MAF MbpcM MethodLOD
χ2
p-value
tsc0592063
rs936232
rs1031739
rs2046015
rs0965823
rs1378876
rs960345
rs0724156
rs1365372
Affy
Illum
Illum
Illum
Affy
Illum
Illum
Affy
Illum
0.45
0.42
0.40
0.43
0.49
0.37
0.46
0.24
0.12
13.984
25.564
87.688
94.144
96.436
97.354
105.689
131.384
139.914
27.3
37.1
93.1
98.1
102.3
100.4
105.6
128.1
129.8
QPDT
REG
QPDT
QPDT
QPDT
REG
QPDT
QPDT
REG
-a
11.38
-
6.99
7.47
4.35
-
8.50
8.62
-
0.001
-
0.008
0.006
0.037
-
0.004
0.003
-
1.94
-
-
-
2.14
-
-
1.91
a-, not applicable.

Page 4

BMC Genetics 2005, 6:S124
Page 4 of 5
(page number not for citation purposes)
be microsatellites positioned relatively close by. Other
possible explanations for the higher LOD scores in the
SNPs could be higher genotyping success rates in the SNPs
or greater genotyping error in the microsatellites. The dis-
crepancy could also be due to chance, for example if less
informative SNPs resulted in some families being unin-
formative for linkage who might otherwise have contrib-
uted evidence against linkage.
Note also that in the microsatellite screens, the two-point
analysis in many instances found higher LOD scores than
the multipoint analysis. For example, the two-point anal-
ysis found weak evidence on chromosome 7 with a LOD
score over 1 at 116.6 cM (in the vicinity of the SNP linkage
signals), while multipoint LOD scores did not reach 1 on
chromosome 7 (the maximum LOD was 0.89 at 140 cM).
Conclusion
Our two-stage genome-wide analysis implicates chromo-
somes 2 and 7 as potentially harboring loci influencing
the maximum drinks consumption phenotype. Follow-up
association analysis supports chromosome 7 more
strongly than chromosome 2, and suggests that further
fine-mapping efforts on chromosome 7 may be construc-
tive. Interestingly, this region of chromosome 7q contains
several TAS2R bitter taste receptor genes [7], and haplo-
type status at one of these, TAS2R38, has been shown to
be a significant predictor of alcohol intake [8]. The mus-
carinic receptor CHRM2 gene associated with alcohol
dependence and major depressive syndrome in COGA [9]
is also in this region.
Chromosome 4, which has been implicated in a previous
linkage study of maximum drinks, did not pass our
threshold for linkage evidence in the first stage of our pri-
mary analysis, but additional analyses of this chromo-
some indicated modest evidence for both linkage and
association. In Saccone et al. [1] linkage to maximum
drinks was found at marker D4S2407 on chromosome 4,
which is located at 104.75 cM according to the Marshfield
genetic map. Build 35.1 of the NCBI physical map puts
this marker at approximately 97.60 Mb. This is near our
linkage LOD score of 2.14 at rs1378876 located at 97.25
Mb on the same build (97.35 Mb on the GAW Illumina
map). Hence, our supplementary analysis indicates some
support for prior results on chromosome 4; however, our
primary, two-stage analysis of the full genome does not
support chromosome 4 and instead points to chromo-
some 7 as the strongest candidate for harboring a quanti-
tative trait locus. The contrast between the chromosome 4
results for our primary analyses reported here and previ-
ous COGA findings is likely explained in part by the fact
that the sample of families available for GAW14 is a sub-
set of the full COGA dataset that was used in previously
published COGA papers.
LD between markers is a concern when using multipoint
linkage analysis, and our results show there is strong LD
among SNPs on chromosome 7 which were found to be
associated with maximum drinks. As an additional analy-
sis we studied global LD structure using an algorithm we
have previously described [10] to produce a non-overlap-
ping set of blocks across all the chromosomes in the Illu-
mina and Affymetrix SNP sets and found a significant
amount of LD and variable block coverage on the differ-
ent chromosomes, as expected. We are developing tech-
niques for selecting "tag" SNPs from LD blocks which
should be useful in this context by allowing maps to be
thinned for multipoint linkage analysis, or by targeting
"tag" SNPs for priority genotyping.
Abbreviations
COGA: Collaborative Study on the Genetics of Alcohol-
ism
GAW14: Genetic Analysis Workshop 14
LD: Linkage disequilibrium
MAF: Minor allele frequency
NCBI: National Center for Biotechnology Information
QPDT: Quantitative pedigree disequilibrium test
SNP: Single-nucleotide polymorphism
Authors' contributions
SFS performed all statistical analyses and is the author of
the program used to find LD blocks. All authors assisted
in the design of the study and interpretation of results. In
particular, NLS and JPR assisted in the design of the LD
block program. SFS and NLS drafted the manuscript with
input from all authors, and all authors read and approved
the final manuscript.
Acknowledgements
COGA is supported by NIH grant U10AA08403. This work was also sup-
ported by NIAAA training grant AA07580 (SFS), and NIDA grant DA15129
(NLS).
References
1.Saccone NL, Kwon JM, Corbett J, Goate A, Rochberg N, Edenberg HJ,
Foroud T, Li TK, Begleiter H, Reich T, Rice JP: A genome screen of
the maximum number of drinks as an alcoholism phenotype.
Am J Med Genet 2000, 96:632-637.
2. Abecasis GR, Cherny SS, Cookson WO, Cardon LR: Merlin-rapid
analysis of genetic maps using sparse gene flow trees. Nat
Genet 2001, 30:97-101.
3.Monks SA, Kaplan NL: Removing the sampling restriction from
family-based tests of association for a quantitative-trait
locus. Am J Hum Genet 2000, 66:576-592.
4. Dudbridge F: Pedigree disequilibrium test for multilocus hap-
lotypes. Genet Epidemiol 2003, 25:115-121.

Page 5

Publish with BioMed Central and every
scientist can read your work free of charge
"BioMed Central will be the most significant development for
disseminating the results of biomedical research in our lifetime."
Sir Paul Nurse, Cancer Research UK
Your research papers will be:
available free of charge to the entire biomedical community
peer reviewed and published immediately upon acceptance
cited in PubMed and archived on PubMed Central
yours — you keep the copyright
Submit your manuscript here:
http://www.biomedcentral.com/info/publishing_adv.asp
BioMedcentral
BMC Genetics 2005, 6:S124
Page 5 of 5
(page number not for citation purposes)
5.Sham PC, Purcell S, Cherny S, Abecasis R: Powerful regression-
based quantitative-trait linkage analysis of general pedi-
grees. Am J Hum Genet 2002, 71:238-253.
Abecasis GR, Cookson WOC: GOLD-Graphical overview of
linkage disequilibrium. Bioinformatics 2000, 16:182-183 [http://
www.ncbi.nlm.nih.gov/]. National Center for Biotechnology Informa-
tion
Adler E, Hoon MA, Mueller KL, Chandrashekar J, Ryba NJP, Zuker
CS: A novel family of mammalian taste receptors. Cell 2000,
100:693-702.
Duffy VB, Davidson AC, Kidd JR, Kidd KK, Speed WC, Pakstis AJ,
Reed DR, Snyder DJ, Bartoshuk LM: Bitter receptor gene
(TAS2R38), 6-n-probylthiouracil (PROP) bitterness and
alcohol intake. Alcohol Clin Exp Res 2004, 28:1629-1637.
Wang JC, Hinrichs AL, Stock H, Budde J, Allen R, Bertelsen S, Kwon
JM, Wu W, Dick DM, Jones K, Nurnberger JI Jr, Tischfield J, Porjesz
B, Edenberg HJ, Hesselbrock V, Crowe R, Schuckit M, Begleiter H,
Reich T, Goate AM, Bierut LJ: Evidence of common and specific
genetic effects: association of the muscarinic acetylcholine
receptor M2 (CHRM2) gene with alcohol dependence and
major depressive syndrome. Hum Mol Genet 2004,
13:1903-1911.
Taillon-Miller P, Saccone SF, Saccone NL, Duan S, Kloss EF, Lovins
EG, Donaldson R, Phong A, Ha C, Flagstad L, Miller S, Drendel A, Lind
D, Miller RD, Rice JP, Kwok P-Y: Linkage disequilibrium maps
constructed with common SNPs are useful for first-pass dis-
ease association screens. Genomics 2004, 84:899-891.
6.
7.
8.
9.
10.