Application of family-based association testing to assess the genotype-phenotype association involved in complex traits using single-nucleotide polymorphisms.

Abstract

We used the FBAT (family-based association test) software to test for association between 300 individual single-nucleotide polymorphisms and P1 (a latent trait of Kofendred Personality Disorder) in 100 simulated replicates of the Aipotu population. Using the Genetic Analysis Workshop 14 dataset, we calculated the power of FBAT to detect linkage disequilibrium on chromosome 3 (D2). Also, we calculated the false-positive rate on chromosome 1, which contains a true locus (D1) but no linkage disequilibrium was simulated between the trait and all the surrounding single-nucleotide polymorphisms.
We were able to detect the associations between phenotype P1 and three adjacent markers B03T3056 (average p-value = 0.0002), B03T3057 (average p-value = 0.00072), and B03T3058 (average p-value = 0.0038) with power of 98%, 87%, 71% on chromosome 3, respectively. The overall false positive rate to detect association was 0.06 on chromosome 1.
The power to detect a significant association in 100 nuclear families affected with the latent trait of Kofendred Personality Disorder by using FBAT was reasonable (based on 100 replicates). In the future, we will compare the performance of FBAT with alternative approaches, such as using FBAT-generalized estimating equations methods to test for association in families affected with complex traits.

Full text

Page 1

BioMed Central
Page 1 of 4
(page number not for citation purposes)
BMC Genetics
Open Access
Proceedings
Application of family-based association testing to assess the
genotype-phenotype association involved in complex traits using
single-nucleotide polymorphisms
Ming-Hsi Wang1, Mitchell Guo2 and Yin Y Shugart*1
Address: 1Department of Epidemiology, Bloomberg School of Public Health, 615 North Work Street, Johns Hopkins University, Baltimore,
Maryland 21205, USA and 2University of Maryland, College Park, Maryland, USA
Email: Ming-Hsi Wang - mwang2@jhsph.edu; Mitchell Guo - rukawa110@gmail.com; Yin Y Shugart* - yyao@jhsph.edu
* Corresponding author
Abstract
Background: We used the FBAT (family-based association test) software to test for association
between 300 individual single-nucleotide polymorphisms and P1 (a latent trait of Kofendred
Personality Disorder) in 100 simulated replicates of the Aipotu population. Using the Genetic
Analysis Workshop 14 dataset, we calculated the power of FBAT to detect linkage disequilibrium
on chromosome 3 (D2). Also, we calculated the false-positive rate on chromosome 1, which
contains a true locus (D1) but no linkage disequilibrium was simulated between the trait and all the
surrounding single-nucleotide polymorphisms.
Results: We were able to detect the associations between phenotype P1 and three adjacent
markers B03T3056 (average p-value = 0.0002), B03T3057 (average p-value = 0.00072), and
B03T3058 (average p-value = 0.0038) with power of 98%, 87%, 71% on chromosome 3,
respectively. The overall false positve rate to detect association was 0.06 on chromosome 1.
Conclusion: The power to detect a significant association in 100 nuclear families affected with the
latent trait of Kofendred Personality Disorder by using FBAT was reasonable (based on 100
replicates). In the future, we will compare the performance of FBAT with alternative approaches,
such as using FBAT-generalized estimating equations methods to test for association in families
affected with complex traits.
Background
For complex diseases such as Kofendred Personality Dis-
order (KPD), linkage analysis using microsatellite markers
may not be able to provide adequate resolution to identify
the genes underlying phenotypic variation [1]. Fine map-
ping of those linked regions may be accomplished by
using joint tests for linkage and association [2]. Family-
based association tests (FBAT) [3] are a positional
genomic strategy that can test for association in areas with
identified linkage and can be used as a tool to detect asso-
ciation in candidate gene regions with no previously
detected linkage signals [4].
The aim of this study is to use FBAT to test for association
between single-nucleotide polymorphism (SNP) markers
and the P1 phenotype (a latent KPD trait containing 4 dif-
ferent phenotypes, i.e., fear/discomfort with strangers,
humor impairment, fascination with automobiles, and
uncommunicative speech patterns) using SNPs on chro-
mosomes 1 and 3. We evaluated the power to detect asso-
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):S68doi:10.1186/1471-2156-6-S1-S68
<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:S68
Page 2 of 4
(page number not for citation purposes)
ciation using FBAT in a simulated dataset of 100 replicates
of the Aipotu population.
Methods
FBAT has been used to test for genetic association by some
investigators [5,6]. It builds on the original transmission-
disequilibrium test proposed by Ewens and Spielman [7],
in which alleles transmitted to affected offspring are com-
pared with the expected distribution of alleles among off-
spring. Moreover, it offers options to test for association in
the presence of linkage or without linkage, using either
single SNPs or haplotypes. Laird et al. [4] proposed to use
an empirical variance-covariance estimator that adjusts
for the correlation among siblings' marker genotypes and
for different nuclear families within the same extended
pedigree as a validity test for the association between
marker and disease status. Because FBAT uses these condi-
tional distributions in deriving the distribution for the test
statistic under the null hypothesis, biases due to popula-
tion admixture, misspecification of the trait distribution,
and/or selection based on trait can be avoided.
Our goal is to test the hypothesis of no association using
genotype data in 100 nuclear families, each with different
sibship size, provided by Genetic Analysis Workshop 14
(GAW14). We focused on two regions with known disease
loci: chromosomes 1 and 3. For chromosome 1, we ana-
lyzed 230 SNPs (with average density of 0.3 cM), covering
the region from 117 cM to 191 cM) containing the true
disease locus D1, located at 167 cM. For chromosome 3,
we analyzed 84 SNPs with the same average SNP density
as chromosome 1 (covering the region from 274 cM to
299 cM), and containing the true disease locus D2,
located at 299 cM.
As described by Greenberg et al. [8], Aipotu families were
selected when at least two offspring were present who had
the P1 latent trait and other family members were coded
as "affected" if they were diagnosed with P1.
Results
Power calculation
For the latent trait P1, the average p-value for SNPs on
chromosome 1 over 100 replicates was not significant
(Figure 1A). For chromosome 3, the average p-value was
always greater than 0.05 except for SNPs B03T3056,
B03T3057, and B03T3058, which had average p-values of
0.00002, 0.00074, and 0.0038, respectively, showing
highly significant evidence for association (Figure 1B).
These three adjacent SNPs were approximately located at
position 296 cM on chromosome 3 (within the simulated
LD region, covering 3 cM between B03T3056 and
B03T3067).
Furthermore, we calculated the power to detect associa-
tion between markers and the P1 latent trait with FBAT. If
we defined a significant p-value to be less than 0.05, the
power to detect a significant association was 98%, 87%,
71% for SNPs B03T3056, B03T3057, and B03T3058,
respectively. The highest power was detected at
B03T3056, which is situated in the designated linkage dis-
equilibrium (LD) region of chromosome 3 and located
2.3 cM proximal to the "true" disease locus D2 (Figure
2B).
Calculating the proportion of SNPs giving p-values less
than 0.05
We also calculated the number of SNPs on chromosome
1, in which no association with the trait was simulated,
that would meet the significance threshold of 0.05. We
tested all 230 SNPs on chromosome 1 individually in 100
replicates using FBAT and then we calculated the propor-
tion of significant markers among all tested SNPs using
different cut-off p-values (Figure 2A). First, we counted the
number of SNPs giving p-values less than 0.05 in each rep-
licate and summed them over all 100 replicates. The total
sum over all 230 SNPs that gave a p-value less than 0.05
was 1,374. Then, we divided the sum by the total number
of tests performed. Although we wished to conduct tests
on all 230 SNPs and 100 replicates, some SNPs had an
insufficient number of informative families for FBAT to
calculate the test statistic. We therefore performed fewer
Average p-value of 100 replicates for each SNP on chromo-some 1 (A) and chromosome 3 (B)
Figure 1
Average p-value of 100 replicates for each SNP on chromo-
some 1 (A) and chromosome 3 (B).

Page 3

BMC Genetics 2005, 6:S68
Page 3 of 4
(page number not for citation purposes)
tests (22,802) than the maximum possible (23,000). The
estimated false-positive rate was 0.06. The proportion of
SNPs out of 100 replicates p-values less than 0.05 is pre-
sented in detail in Figure 2A. The average individual p-val-
ues are described in Figure 1A.
Discussion and conclusion
For complex diseases, such as KPD here, we need new sta-
tistical tools such as FBAT to detect associations between
marker loci and disease genes where the disease pheno-
type is multivariate. In this study, we used 100 simulated
replicates of the Aipotu population to calculate the power
to detect association and evaluate the false-positive rate.
We would like to point out two limitations of this study.
We were interested in testing for association with a latent
trait containing 4 phenotypes. Therefore, we conducted
multivariate analysis using FBAT. Given the fact that we
used 100 nuclear families each with a single sibship to test
association, a more appropriate method would have been
to use the "-e" option implemented in FBAT to calculate
the empirical variance of the test statistic to test for associ-
ation in the presence of linkage [2,9]. However, the "-e"
option is not implemented for multivariate analysis using
the current version of FBAT. We recognize that under the
null hypothesis of "no association in the presence of link-
age", different nuclear families within the same pedigree
cannot be treated independently, and furthermore trans-
missions to different sibs in the same nuclear family can-
not be treated as independent. In our study, we analyzed
100 nuclear families with an average number of 4.8 sibs
per pedigree and the use of the "-e" option is desirable.
However, based on the description given in the FBAT tuto-
rial kit (available online at http://www.biostat.har
vard.edu/~clange/default.htm), the results obtained by
using "-e" to test genetic association do not differ greatly
from the result obtained from not using "-e" in nuclear
families, unless there are a few very large pedigrees that
contribute most of the information. In addition, when we
calculated the false-positive rate, we did not take into
account the fact that some SNPs are correlated. These two
limitations could bias the estimated false-positive rates. In
this study, if we set the significance level to be 0.05, the
proportion of observed "significant" results was 6%,
which is slightly higher than the expected 5%. However,
given the limitation we discussed above, we cannot con-
clude that this result suggests an inflated type I error.
To summarize, our results indicated the best power of
98% at the SNP B03T3056, within the designated LD
region of chromosome 3, and for adjacent markers
B03T3057 and B03T3058, the power was 87% and 71%,
respectively. None of the other markers within the desig-
nated LD region revealed significant results. We conclude
that FBAT provides another powerful approach to detect
association in the presence of linkage.
Abbreviations
FBAT: Family-based association tests
KPD: Kofendrerd Personality Disorder
LD: Linkage disequilibrium
SNP: Single-nucleotide polymorphism
Authors' contributions
YYS and M-HW both participated in study design and data
analyses. MG and M-HW provided bioinformatics sup-
port to speed up the data analyses. YYS, M-HW, and MG
contributed to data interpretation and manuscript prepa-
rations.
Acknowledgements
YYS is partially supported by the Department of Epidemiology at the Johns
Hopkins Bloomberg School of Public Health and by RO3 CA113240-01.
We would like to thank three anonymous reviewers for their thoughtful
criticism and thank Dr. Xin Xu for his helpful comments.
References
1.Risch N, Merikangas K: The future of genetic studies of complex
human diseases. Science 1996, 273:1516-1519.
2.Lake SL, Blacker D, Laird NM: Family-based tests of association
in the presence of linkage. Am J Hum Genet 2000, 67:1515-1525.
Proportion of 100 replicates showing significant p-value in each SNP at chromosome 1 (A) chromosome 3 (B)
Figure 2
Proportion of 100 replicates showing significant p-value in
each SNP at chromosome 1 (A) chromosome 3 (B).

Page 4

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:S68
Page 4 of 4
(page number not for citation purposes)
3. Rabinowitz D, Laird NM: A unified approach to adjusting asso-
ciation tests for population admixture with arbitrary pedi-
gree structure and arbitrary missing marker information.
Hum Hered 2000, 50:211-223.
Laird NM, Horvath S, Xu X: Implementing a unified approach to
family-based tests of association. Genet Epi 2000, 19:S36-S42.
Horvath S, Xu X, Lake SL, Silverman EK, Weiss ST, Laird NM: Fam-
ily-based tests for associating haplotypes with general phe-
notype data: application to asthma genetics. Genet Epidemiol
2004, 26:61-69.
Skaar DA, Shao Y, Haines JL, Stenger JE, Jaworski J, Martin ER, Delong
GR, Moore JH, McCauley JL, Sutcliffe JS, Ashley-Koch AE, Cuccaro
ML, Folstein SE, Gilbert JR, Pericak-Vance MA: Analysis of the
RELN gene as a genetic risk factor for autism. Mol Psychiatry .
2004, Nov 23; doi:10.1038/sj.mp.4001614
Ewens W, Spielman RS: The transmission/disequilibrium test:
history, subdivision and admixture. Am J Hum Genet 1995,
57:455-464.
Greenberg DA, Zhang J, Shmulewitz D, Strug LJ, Zimmerman R, Singh
V, Marathe S: Construction of the model for the Genetic Anal-
ysis Workshop 14 simulated data: genotype-phenotype rela-
tionships, gene interaction,
disequilibrium, and ascertainment effects for a complex phe-
notype. BMC Genet 2005, 6(Suppl 1):S3.
Lange C: A multivariate family-based association test using
generalized estimating equation: FBAT-GEE. Biostatistics
2003, 4:195-206.
4.
5.
6.
7.
8.
linkage, association,
9.