%% \VignetteIndexEntry{gap: genetic analysis package}
%% \Vignettekeywords{genetic data analysis}
%% \VignettePackage{gap}
\documentclass[article,shortnames]{jss}
\usepackage{thumbpdf,graphicx}
\author{Jing Hua Zhao\\MRC Epidemiology Unit}
\Plainauthor{Jing Hua Zhao}
\title{\pkg{gap}: Genetic Analysis Package}
\Plaintitle{gap: Genetic Analysis Package}
\Abstract{A preliminary attempt at collecting tools and utilities
for genetic data as an \proglang{R} package called \pkg{gap} is
described. Genomewide association is then described as a specific
example, linking the work of \cite{risch96}, \cite{ll97} for
family-based and population-based studies, and the counterpart for
case-cohort design established by \cite{cz04}. Analysis of staged
design as outlined by \cite{skol06} and associate methods are
discussed. The package is flexible, customizable, and should prove
useful to researchers especially in its application to genomewide
association studies.}
\Keywords{genetic data analysis, genomewide association, \proglang{R}}
\Plainkeywords{genetic data analysis, genomewide association, R}
\Volume{23}
\Issue{8}
\Month{December}
\Year{2007}
\Submitdate{2007-01-03}
\Acceptdate{2007-06-08}
\Address{
Jing Hua Zhao\\
MRC Epidemiology Unit\\
Institute of Metabolic Science\\
Addenbrooke's Hospital\\
Hills Road\\
Cambridge CB2 0QQ\\
United Kingdom\\
E-mail: \email{jinghua.zhao@mrc-epid.cam.ac.uk}\\
}
\begin{document}
\section{Introduction}
Approaches to understanding the genetic basis of human diseases have
been widely discussed, e.g., \cite{morton83}, \cite{khoury93},
\cite{thomas04}. Methods include the assessment of familial
aggregation for heritability, identification of major gene effect,
study of cosegregation of genetic marker with putative
disease-predisposing loci in the so-called linkage studies and
association studies in search of frequency differences between cases
and controls and/or correlation between genotype and phenotype as a
quantitative trait. Recently, owing to the availability of large
number of genetic variants and particularly single nucleotide
polymorphisms (SNPs), attention has focused on association designs
including both families and unrelated individuals from general
populations. Three initiatives of interest are:
\begin{enumerate}
\item The hapmap project (\url{http://www.hapmap.org/}), a
partnership of scientists and funding agencies from Canada, China,
Japan, Nigeria, the United Kingdom and the United States to develop
a public resource that will help researchers find genes associated
with human disease and response to pharmaceuticals.
\item The Wellcome Trust Case Control Consortium (WTCCC,
\url{http://www.wtccc.org.uk/}), a collaboration of human
geneticists who analyse thousands of DNA samples from patients
suffering with different diseases to identify common genetic
variations for each condition. It is hoped that by identifying these
genetic signposts, researchers will be able to understand which
people are most at risk, and also produce more effective treatments.
The WTCCC searches for the genetic signposts for tuberculosis,
coronary heart disease, type 1 diabetes, type 2 diabetes, rheumatoid
arthritis, Crohn's disease, bipolar disorder and hypertension. The
research is conducted at a number of institutes throughout
the UK, including the Wellcome Trust Sanger Institute, Cambridge
University and Oxford University.
\item The genetic association information network
(GAIN, \url{http://www.fnih.org/GAIN2/home_new.shtml}), a
public-private partnership of the Foundation for the National
Institutes of Health, Inc., which include corporations, private
foundations, advocacy groups, concerned individuals, and the
National Institutes of Health. This initiative will take the next
step in the search to understand the genetic factors influencing
risk for complex human diseases.
\end{enumerate}
Following successes in localization of Mendelian diseases such as
Huntington's disease \citep{guesella83}, cystic fibrosis
\citep{tsui85}, some recent successes in non-Mendelian diseases come
from breast cancer \citep{hall90}, macular degenration
\citep{klein05}, and non-insulin-dependent (Type-2) diabetes
\citep{grant06}. Here, we will focus specifically on methodological
issues concerning the design and analysis of genomewide association
studies as largely foreseen by the seminal paper of \cite{risch96}.
As it dealt with the power of association studies using family-based
designs, an immediate argument for case-control designs as an
alternative to family-based designs was made by \cite{ll97} on the
basis of (1). its ease of implementation; (2). the increased
prospects for extension from established epidemiological cohorts;
(3). for late onset diseases such as Type-2 diabetes and
hypertension, the difficulty of typing parents for family-based
studies; (4). the issue that, for equivalent power, the number of
individuals genotyped needs to be doubled for linkage, tripled for
singleton, and quadrupled for sib-pair designs, assuming both
parents are genotyped in an affected sibling study. Concerns over
cost-efficiency have led to the adoption of staged design in some
studies. In the most popular two-staged design, the first stage uses
only a proportion of individuals and are genotyped at all SNPs,
among which a percentage of SNPs showing statistical significance is
carried over to the second stage. For analysis of two-staged design,
\cite{skol06} recognized that joint analysis could be more powerful.
The high cost of genotyping has motivated researchers to seek
generic controls for a range of traits as in a case-cohort design,
in which a small random sample of the whole cohort and all the
diseased subjects are used. It is possible the subcohort also
contains some cases but it is fully representative of the population
and can be used in conjunction with a range of case definitions. The
merits of such a design in genetic association studies have only
recently been recognized \citep{langholz99,manolio06}. Some of the
terms used in this paper can be found in the paper by \cite{bgm06},
and more generally in a review paper by \cite{ea06} on recent
developments in statistical genetics, and in a tutorial specifically
for genetic association analysis by \cite{balding06}.
Analysis for population association studies generally involves
preliminary analyses such as quality control, Hardy-Weinberg
equilibrium tests, examination of linkage disequilibrium and
recombination, to be followed by tests of association for single
and/or multiple SNPs, both may involve case-control or binary
phenotype or continuous outcomes. Further issues involve handling of
large data, multiple testing, among others. The implementation of
analytical approaches to these problems has so far scattered and not
integrated with established software systems. Preliminary reviews
have shown that \proglang{R} is a strong alternative which offers
many desirable facilities and can be used in conjunction with other
software systems \citep{zt06b}. It is possible that in the near
future, many tools for genetic data analysis will be available on
these software systems.
We have recently carried out a study of obesity using Affymetrix
(\url{http://www.affymetrix.com/}) 500K and Illumina
(\url{http://www.illumina.com/}) 317K systems following an earlier
pilot using Perlegen (\url{http://www.perlegen.com/}) platform with
about 250K SNPs. The samples were based on European Prospective
Investigation of Cancer (EPIC) Norfolk study
(\url{http://www.srl.cam.ac.uk/epic/}). Our particular concern is
cost-efficiency, not only because the experiment is expensive but we
wish to take full advantage of it. We adopted a two-stage
case-cohort design which is advantageous in cost-efficiency over the
standard case-control design. While seeking for appropriate methods
for joint analyses, we have recast the problem within the missing
data framework. We believe our approach will be invaluable to other
colleagues.
Here, I describe a genetic analysis package, \pkg{gap}, created and
maintained by the author and is available from the Comprehensive
\proglang{R} Archive Network (CRAN,
\url{http://CRAN.R-project.org/}), taking advantages of the portable
environment of \proglang{R} for data management, analysis, graphics
and object-oriented programming. The functions implemented for
genetic association analysis include Hardy-Weinberg equilibrium
tests, measures of linkage disequilibrium between SNPs or
multiallelic markers and haplotype analysis. I will give a brief
overview of the package, and a specific example concerning design
and analysis of genetic association, which was used in the design of
the case-cohort study aforementioned. I will also use a dataset from
our case-control association study of Type-2 diabetes with
chromosome 20, to illustrate joint analysis within \proglang{R}. It
is clear that many statistical issues can be handled with ease and
used in conjunction with other packages available from CRAN. The
calculation as in some original work is also given for reference,
with results not seen in the genetic analysis literature. At the end
I will provide a brief summary.
\section{Overview of implemented methods}
This package was designed in 2003 to integrate some programs in
\proglang{C}/\proglang{Fortran} and \proglang{SAS} \citep{SASstat} I
have written or used over the years, by taking advantage of the
ability of \proglang{R} to use foreign language routines in
\proglang{C}/\proglang{Fortran} or its facilities similar to
\proglang{SAS}. The recent implementations are more native to
\proglang{R}. The collection therefore somewhat reflects the
evolution of (or more exactly my appreciation of) statistical
genetics (or genetic epidemiology) over years. The description of
the main functions is given in Table~\ref{pkggap}, while a list of
data sets is given in Table~\ref{dataset}. They can be conveniently
obtained with \proglang{R} command \code{library(help = "gap")} once
\pkg{gap} is installed.
\begin{table}
\begin{center}
\begin{tabular}{ll}
\hline
Function name & Description\\
\hline
\texttt{BFDP} & Bayesian false-discovery probability \\
\texttt{FPRP} & False-positive report probability\\
\texttt{bt} & Bradley-Terry model for contingency table\\
\texttt{ccsize} & Power and sample size for case-cohort design\\
\texttt{chow.test}& Chow's test for heterogeneity in two regressions\\
\texttt{fbsize} & Sample size for family-based linkage and association design\\
\texttt{gc.em} & Gene counting for haplotype analysis\\
\texttt{gcontrol} & genomic control\\
\texttt{gcp} & Permutation tests using \pkg{GENECOUNTING}\\
\texttt{genecounting} & Gene counting for haplotype analysis\\
\texttt{gif} & Kinship coefficient and genetic index of familiality\\
\texttt{hap} & Haplotype reconstruction\\
\texttt{hap.em} & Gene counting for haplotype analysis\\
\texttt{hap.score}& Score statistics for association of traits with haplotypes\\
\texttt{htr} & Haplotype trend regression\\
\texttt{hwe} & Hardy-Weinberg equilibrium test for a multiallelic marker\\
\texttt{hwe.hardy} & Hardy-Weinberg equilibrium test using MCMC\\
\texttt{kbyl} & Linkage disequilibrium statistics for two multiallelic loci\\
\texttt{kin.morgan} & Kinship matrix for simple pedigree\\
\texttt{makeped} & A function to prepare pedigrees in post-\pkg{MAKEPED} format\\
\texttt{mia} & Multiple imputation analysis for hap\\
\texttt{mtdt} & Transmission/disequilibrium test of a multiallelic marker\\
\texttt{muvar} & Means and variances under 1- and 2- locus (biallelic) QTL model\\
\texttt{pbsize} & Power for population-based association design\\
\texttt{pedtodot} & Converting pedigree(s) to \texttt{dot} file(s)\\
\texttt{pfc} & Probability of familial clustering of disease\\
\texttt{pfc.sim} & Probability of familial clustering of disease\\
\texttt{pgc} & Preparing weight for \pkg{GENECOUNTING}\\
\texttt{plot.hap.score} & Plot haplotype frequencies versus haplotype score statistics\\
\texttt{print.hap.score}& Print a \texttt{hap.score} object\\
\texttt{s2k} & Statistics for 2 by K table\\
\texttt{tbyt} & Linkage disequilibrium statistics for two SNPs\\
\texttt{tscc} & Power calculation for two-stage case-control design\\
\texttt{twinan90} & Classic twin models\\
\texttt{whscore} & Whittemore-Halpern scores for allele-sharing\\
\hline
\end{tabular}
\caption{\label{pkggap} A list of functions in \pkg{gap}.}
\end{center}
\end{table}
\begin{table}
\begin{center}
\begin{tabular}{ll}
\hline
Data sets & Description\\
\hline
\texttt{aldh2} & ALDH2 markers and alcoholism\\
\texttt{apoeapoc} & APOE/APOC1 markers and schizophrenia\\
\texttt{cf} & Cystic fibrosis data\\
\texttt{crohn} & Crohn's disease data\\
\texttt{fa} & Friedreich ataxia data\\
\texttt{fsnps} & A case-control data involving four SNPs with missing genotype\\
\texttt{hla} & HLA markers and schizophrenia\\
\texttt{mao} & A study of Parkinson's disease and MAO gene\\
\texttt{nep499} & A study of Alzheimer's disease with eight SNPs and APOE\\
\texttt{snca} & A study of Parkinson's disease and SNCA markers\\
\hline
\end{tabular}
\caption{\label{dataset} A list of data sets in \pkg{gap}.}
\end{center}
\end{table}
Some of the functions are highlighted as follows:
\begin{itemize}
\item \texttt{BFDP}: Bayesian false-discovery probability according to \cite{wakefield07}.
\item \texttt{FPRP}: False-positive report probability according to
\cite{wacholder04}.
\item \texttt{gif}: kinship coefficient and genetic index of familiality according to \cite{gholamic94}.
\item \texttt{genecounting}, \texttt{gcp}, \texttt{gc.em}, \texttt{pgc}: expectation-maximization (EM) method
to infer haplotype and case-control haplotype association similar to
\proglang{SAS}/\pkg{Genetics} \citep{sasgenetics} as reported in
\cite{zhao02}, \cite{zhao04}.
\item \texttt{hap}, \texttt{hap.em}, \texttt{mia}: a multiallelic version of progressive EM algorithm for haplotype inference as reported in \cite{zhao04}.
\item \texttt{tbyt}, \texttt{kbyl}: linkage disequilibrium statistics between two
biallelic or multiallelic markers, including standard error for $D'$
as reported in \cite{zhao04}.
\item \texttt{pfc}, \texttt{pfc.sim}: probability of familial clustering of diseases according to \cite{yz02}.
\item \texttt{htr}: haplotype-trend regression as accorded to \cite{zaykin02}.
\item \texttt{hap.score}, \texttt{print.hap.score}, \texttt{plot.hap.score}: score statistics
for haplotype-trait association with revised algorithms according to
\cite{srt02}.
\item \texttt{pbsize}, \texttt{fbsize}, \texttt{ccsize}: sample
size/power for population-based, family-based case-control designs,
or population-based case-cohort design, according to \cite{ll97},
\cite{risch96}, \cite{cz04}.
\item \texttt{s2k}: statistics from $2\times K$ table according
to \cite{hirotsu01}.
\item \texttt{hwe}, \texttt{hwe.hardy}: Hardy-Weinberg equilibrium test for
multiallelic markers, the latter implements a Markov chain Monte
Carlo (MCMC) algorithm according to \cite{guo92}.
\item \texttt{makeped}: a utility to prepare for post-\pkg{MAKEPED} format file as
required by \pkg{LINKAGE} \citep{lathrop84} or other computer
programs, a version distributed with \pkg{LINKAGE} by Wentian Li.
\item \texttt{pedtodot}: a utility to produce \pkg{Graphviz}
\citep{graphviz} command file based on a \proglang{GAWK} script by
David Duffy.
\item \texttt{tscc}: power of two-stage case-control
design under several models, similar to \cite{skol06}.
\item \texttt{muvar}, \texttt{whscore}: utility functions to obtain mean and variance
under two-locus model, and score \citep{whittemore94b}.
\end{itemize}
A brief summary of a given function, including appropriate
reference(s), can be obtained with \code{help(function-name, package
= "gap")} within \proglang{R}. There are a number of example
datasets. For instance, the cystic fibrosis data \citep{kerem89} has
been used in many papers on fine-mapping.
\section{Example: Study design and analysis of genetic association}
Since the range of problems to be tackled by \pkg{gap} is quite
broad, I now give specific examples of how some functions have been
developed to facilitate our design and analysis of genetic
association. The three popular designs outlined earlier are shown in
Figure~\ref{3design}.
\begin{figure}
\begin{center}
\includegraphics{figure1.jpg}
\caption{\label{3design} Three common genetic association designs involving
unrelated individuals (left), nuclear families with affected
singletons (middle) and affected sib-pairs (right). Males and
females are denoted by squares and circles with affected individuals
filled with black and unaffect individuals being
empty.}
\end{center}
\end{figure}
Because of the importance of work by \cite{risch96} and \cite{ll97},
functions \texttt{pbsize} and \texttt{fbsize} have been written to
implement their methods and correct a programming error in
\cite{risch96}. The table to be obtained could be used as a quick
reference and the code be tailored to particular designs. Power for
case-cohort design established by \cite{cz04} is given in function
\texttt{ccsize}. The method by \cite{skol06} for two-stage design
has been implemented in function \texttt{tscc}.
\subsection{Some preliminaries}
A widely used general model consists of a disease susceptible locus
with two alleles (A and a) and population frequencies $p$, $q=1-p$,
and penetrances as $f_0$, $f_1$ and $f_2$ (i.e., the conditional
probability of getting the disease, the subscripts 0, 1 and 2 are
copies of the disease-predisposing allele A). Hardy-Weinberg
equilibrium states that the distribution of the three genotypes, aa,
Aa, and AA are according to $(p+q)^2$, so that the population
prevalence of the disease $K = p^2f^2+2pqf_1+q^2f_0$. Other
statistics, which are often used, are additive $V_A =
2pq[p(f_2-f_1)+q(f1-f_0)]^2$ and dominance $V_D =
p^2q^2(f_2-2f_1+f_0)^2$, variances, respectively. It is also common
to specify further the relationship between the three genotypes. As
for multiplicative models to be considered here, the genotypic
relative risk (GRR) is $\gamma$ and $\gamma^2$ for genotypes Aa and
AA compared to aa as baseline, the population disease prevalence,
the additive and dominance variances are therefore
$K=p^2\gamma^2+2pq\gamma+q^2=(p\gamma+q)^2$,
$V_A=2pq(\gamma-1)^2(p\gamma+q)^2$ and $V_D=p^2q^2(\gamma-1)^4$,
respectively. These can be used to define offspring/sibling relative
risks, which are the probability ratios of a offspring/sibling of an
affected parent/child being affected relative to the general
population and they are shown to be $\lambda_O=1+{1/2V_A}/{K^2}$ and
$\lambda_S=1+{(V_A/2+V_D/4)}/{K^2}$ so $\lambda_O=1+w$ and
$\lambda_S=(1+1/2w)^2$, where
$w={(pq(\gamma-1)^2)}/{(p\gamma+q)^2}$.
Linkage and association designs using family data were the
state-of-the-art designs for localising disease-predisposing loci. A
key concept relates to genes shared by relatives from common
ancestor and is called identity by descent (IBD), which is
associated with a model-free approach of linkage whereby
allele-sharing between affected members in a pedigree is compared to
test for linkage between a marker locus and a disease locus. The
simplest allele-sharing method uses affected sib-pairs. Assuming the
recombination fraction between the disease locus and a marker to be
$\theta$, and defining $\Psi = \theta^2+(1-\theta)^2$,
\cite{suarez78} derived the IBD probabilities for affected sib-pair,
$P(IBD=0) = {1}/{4}- [(\Psi-1/2)V_A+(2 \Psi - \Psi^2 - 3/4)V_D] /
[4(K^2 + V_A/2 + V_D/4)]$, $P(IBD=1) = {1}/{2}-
[2(\Psi^2-\Psi+1/4)V_D]/ [4(K^2 + V_A/2 + V_D/4)]$, $P(IBD=2) =
{1}/{4}+ [(\Psi-1/2)V_A+(\Psi^2 - 1/4)V_D] / [4(K^2 + V_A/2 +
V_D/4)]$, respectively. Under no linkage, the probabilities of
affected sib pair sharing 0, 1, and 2 alleles IBD are 1/4, 1/2, and
1/4. This result can be used to form a nonparametric test for
linkage. The corresponding expressions under multiplicative model
can also be obtained, where probabilities of siblings sharing none
or one allele by descent is $z_0=1/(4\lambda_S)$ and
$z_1={\lambda_O}/(2\lambda_S)$, the nonshared probability is
$Y=1-z_1/2-z_0=1-{(\lambda_O+1)}/(4\lambda_S)={(1+w)}/{(2+w)}$. When
a candidate or dense collection of markers is available, one can use
nuclear families in which affected offsprings and their parents are
genotyped, the transmitted versus nontransmitted alleles from
parents to offsprings are compared in the so-called transmission
disequilibrium test (TDT), which can be based on nuclear families
with affected singletons only.
In the following standard results related to normal distribution
will be used. For a set of $N$ independent identically distributed
random variables $B_i$ with mean and variance being 0 and 1 under
the null hypothesis, $\mu$ and $\sigma^2$ under the alternative
hypothesis, the statistic $\sum_{i=1}^N B_i/N$ has mean 0 and
variance 1 under the null but mean $\sqrt{N}\mu$ and variance
$\sigma^2$ under the alternative. The sample size ($N$) for a given
significance level $\alpha$ and power $1-\beta$ can be estimated by
$(Z_\alpha-\sigma Z_{1-\beta})^2/\mu^2$. In the actual calculation
below, the type I error rate ($\alpha$) and type II error rate
($\beta$) are $5\times 10^{-8}$ and 0.2, respectively.
\subsection{Power/sample size for family and case-control/case-cohort designs}
We recall some results on affected sib-pair linkage analysis, and
define the alleles shared and nonshared from the $i$th parent as a
random variable $B_i, i=1,\ldots,N$, scoring $1$ and $-1$. The mean
($\mu$) and variance ($\sigma^2$) of $B_i$ are 0 and 1 under the
null hypothesis as the shared and nonshared each has probability
0.5, and $2Y-1$ and $4Y(1-Y)$ under the alternative. Assuming
sharing of alleles from both parent to be independent, the required
sample size ($N$) for affected sib-pair under $\theta=0$ and no
linkage disequilibrium is ${(Z_\alpha-\sigma
Z_{1-\beta})^2}/{2\mu^2}$, where $Y$ and $w$ are defined as before.
As with TDT, we assume that the disease locus and a nearby locus are
in complete disequilibrium, the number of transmissions of allele A
are scored from heterozygous parents, where the probability ($h$) of
a parent of an affected child being heterozygous is given by
${pq(\gamma+1)}/{(p\gamma+q)}$ and ${pq(\gamma+1)^2}/{[2(\gamma
p+q)^2+pq(\gamma-1)^2]}$ for singletons and sib-pairs, respectively.
For singletons, a random variable $B_i$ is defined taking values
$1/\sqrt{h}$ if parent is heterozygous and transmits A, $0$ if
parent is homozygous, $-1/\sqrt{h}$ if parent is heterozygous and
transmits a. The mean and variance of $B_i$ are 0 and 1 under the
null hypothesis, $\sqrt{h}{(\gamma-1)}/{(\gamma+1)}$ and
$1-h[{(\gamma-1)}/{(\gamma+1)}]^2$ under the alternative. When
sib-pairs instead of singletons are used in TDT analysis, the same
formula for sample size calculation can be applied and the required
number of families is half the expected number since there are two
independent affected sibs.
The results of \code{example("fbsize", package = "gap")} are shown
Table~\ref{asptdt}. Column $N_L$ contains the correct calculation
corresponding to the original paper. The Alzheimer's disease model
is based on \cite{sph97}.
\begin{table}[h]
\begin{center}
\begin{tabular}{ccccccccccc}
\hline
&&\multicolumn{2}{c}{Linkage}&&\multicolumn{4}{c}{Association} \\
\cline{3-4} \cline{6-8}
$\gamma$& $p$ &$Y$ & $N_{L}$
& $P_A$& $H_1$ & $N_{\mathit{tdt}}$ & $H_2$ &$N_{\mathit{asp}/\mathit{tdt}}$&$\lambda_o$ & $\lambda_s$ \\
\hline
\\
4.00 & 0.01 & 0.520 & 6400 & 0.800 & 0.048 & 1098 & 0.112 & 235 & 1.08 & 1.09\\
& 0.10 & 0.597 & 277 & 0.800 & 0.346 & 151 & 0.537 & 48 & 1.48 & 1.54\\
& 0.50 & 0.576 & 445 & 0.800 & 0.500 & 104 & 0.424 & 62 & 1.36 & 1.39\\
& 0.80 & 0.529 & 3023 & 0.800 & 0.235 & 223 & 0.163 & 162 & 1.12 & 1.13\\
\\
2.00 & 0.01 & 0.502 & 445839 & 0.667 & 0.029 & 5824 & 0.043 & 1970 & 1.01 & 1.01\\
& 0.10 & 0.518 & 8085 & 0.667 & 0.245 & 696 & 0.323 & 265 & 1.07 & 1.08\\
& 0.50 & 0.526 & 3752 & 0.667 & 0.500 & 340 & 0.474 & 180 & 1.11 & 1.11\\
& 0.80 & 0.512 & 17904 & 0.667 & 0.267 & 640 & 0.217 & 394 & 1.05 & 1.05\\
\\
1.50 & 0.01 & 0.501 & 6942837 & 0.600 & 0.025 & 19321 & 0.031 & 7777 & 1.00 & 1.00\\
& 0.10 & 0.505 & 101898 & 0.600 & 0.214 & 2219 & 0.253 & 941 & 1.02 & 1.02\\
& 0.50 & 0.510 & 27041 & 0.600 & 0.500 & 950 & 0.490 & 485 & 1.04 & 1.04\\
& 0.80 & 0.505 & 101898 & 0.600 & 0.286 & 1663 & 0.253 & 941 & 1.02 & 1.02\\
\\
\multicolumn{3}{c}{Alzheimer's:}\\
\\
4.50 & 0.15 & 0.626 & 163 & 0.818 & 0.460 & 100 & 0.621 & 37 & 1.67 & 1.78\\
\\
\hline
\end{tabular}
\caption{\label{asptdt} Comparison of linkage and association in nuclear
families required for identification of disease gene:
$\gamma$=genotypic risk ratio; $p$=frequency of disease
allele A; $Y$=probability of allele sharing; $N_{L}$=number of ASP
families required for linkage; $P_A$=probability of transmitting
disease allele A; $H_1$, $H_2$=proportions of heterozygous parents;
$N_{\mathit{tdt}}$=number of family trios; $N_{\mathit{asp}/\mathit{tdt}}$=number of ASP.
families}
\end{center}
\end{table}
It turns out that with $\gamma\le 2$, the expected marker-sharing
only marginally exceeds 50\% for any allele frequency ($p$). The use
of linkage would need large sample size, but direct tests of
association with a disease locus itself can still be quite strong
although it may involve large amount of statistical testing of
associated alleles. Clearly, it is most favorable for diseases that
are relatively common, which has important implications for complex
traits.
Now we consider the case-control design with a statistic directly
testing association between marker and disease. Following
\cite{ll97} and supposing we have a randomly ascertained population
sample, the frequencies of the three disease genotypes AA, Aa and aa
in cases are $\pi\gamma^2 p^2$, $2\pi\gamma pq$, and $\pi q^2$,
respectively, where $\pi$ is the ``baseline'' probability that an
individual with $aa$ genotype being affected. Similarly, the three
frequencies in controls are $(1-\pi\gamma^2)p^2$, $2(1-\pi\gamma)pq$
and $(1-\pi)q^2$. A unit $\chi^2$ statistic can be constructed
using Table~\ref{exprop} as $X^2=\sum (O-E)^2/E$, where $E$s
indicate expected unit frequencies ($\pi p(\gamma p+q)^2$, $\pi
q(\gamma p+q)^2$, $p-\pi p(\gamma p+q)^2$ and $q-\pi q(\gamma
p+q)^2$), and the discrepancies between observed and expected
frequencies all have factor $\pi pq(\gamma p+q)(\gamma-1)$ but with
negative sign before the second and the third items. The statistic
$X^2$ has $\chi_1^2$ distribution under the null hypothesis
($\gamma=1$), and noncentral $\chi_{1,\delta}^2$ distribution with
noncentrality parameter $\delta=[\pi pq(\gamma-1)^2]/[1-\pi (\gamma
p+q)^2]$ or $[\gamma^2 p+q-(\gamma p+q)^2]/[1-\pi (\gamma p+q)^2]$
under the alternative ($\gamma>1$). It is equivalent to derive the
power by $Y=\sqrt{X^2}\sim N(\sqrt{\delta},1)$.
$1-\beta=\Phi(-ZC_1)P(|z_2|>C_2,sign(z_1)=sign(z_2))$ and
$P(|z_1|>C_1)P(|z_j|>C_j||z_1|>C_1)$ for replication-based and joint
analyses, respectively. As our primary interest is the power for the
two types of analyses, we implement it in function \texttt{tscc}
whose format is
\begin{verbatim}
tscc(model, GRR, p1, n1, n2, M, alpha.genome, pi.samples, pi.markers, K)
\end{verbatim}
which requires specification of disease model (multiplicative,
additive, dominant, recessive), genotypic relative risk (GRR), the
estimated risk allele frequency in cases ($p_1$), total number of
cases ($n_1$) total number of controls ($n_2$), total number of
markers ($M$), the false positive rate at genome level
($\alpha_\mathit{genome}$), the proportion of markers to be selected
($\pi_\mathit{markers}$, also used as the false positive rate at
stage 1) and the population prevalence ($K$). Note the disease risks
involving the three genotypes for additive, dominant and recessive
models are calculated as $(1,\mathit{GRR},2\cdot\mathit{GRR}-1)$,
$(1,\mathit{GRR},\mathit{GRR})$, $(1,1,\mathit{GRR})$, respectively.
Power for a number of scenarios for two-stage designs can be
obtained with slight modification of example code in the
documentation of \texttt{tscc}. Interestingly, the \proglang{R}
implementation is much shorter than the \proglang{C} program by
\cite{skol06}.
\cite{lin06} recently proposed a method of analysis for two-stage
designs. Like other contributions, the focus was on the markers
genotyped at both stages. Given that in most studies such as the
case-cohort design outlined above, there is a rich collection of
covariate information and perhaps an equally important aspect is to
use all markers and covariates at both stages in a unified analysis.
This amounts to a standard analysis with missing independent
variables with which statistical methods have been well developed,
e.g., \cite{ibrahim90}, \cite{schafer96}, \cite{vbk99}. The packages
for multiple imputation available from CRAN, such as \pkg{cat},
\pkg{mix}, \pkg{norm}, \pkg{pan} \citep{cat, mix, norm, pan},
\pkg{mice} \citep{mice}, \pkg{mitools} \citep{mitools} can be used.
A useful point relates to prospective versus retrospective methods.
The retrospective counterpart requires more attention but far from
clear \citep{ea06}, although the equivalence of retrospective
versus prospective methods is known \citep{pp79,sr04}. It
has been shown \citep{kt00,tcc05} that prospective
likelihood can give biased estimate due to over representation of
cases or the extremes of the traits compared with the general
population, so that a retrospective model of
allele/genotype/haplotype frequencies conditional on the disease
phenotype via logistic model is preferred. The model is applicable
to a wide range of traits and provides unbiased estimates. This is
in contrast to prospective methods \citep{srt02,sm05}.
These two aspects are well illustrated with a concrete example. We
have recently conducted a study of Type-2 diabetes involving 5,013
Ashkenazi and four UK populations using two-staged design and 4,570
SNPs across a 10Mb region of chromosome 20q. A subset of 2,502
individuals were genotyped on these SNPs at stage one, from which
SNPs with significance level 0.1 together and those within regions
of interest (e.g., HNF4A) were further genotyped on remaining sample
at stage two. In the analysis, a meta-analysis was applied to take
into account the heterogeneity between populations. The two-stage
method by \cite{lin06} would be rather complicated. A useful
prospective formulation for a particular marker not genotyped at
stage two is to treat it as missing and considered in a weighted
regression on trait, where the weight is associated with the three
genotypes, similar to \cite{srt02}. For the retrospective method,
one can augment the missing genotype similarly, in much the same way
as \cite{bgm06}. However, recall that this is exactly the kind of
problem multiple imputation will deal with, such that we simply
apply the appropriate packages aforementioned which accounts for the
uncertainty due to missing genotypes. This has not been used in
staged design although it was recognized recently in haplotype
analysis \citep{sorensen06,cordell06}. Now with our data
contained in \texttt{chr20}, and stage one marker rs1419383 is to be
analyzed, we can use \pkg{mice} as follows,
\begin{verbatim}
R> data("chr20")
R> imp <- mice(chr20)
R> fit <- lm.mids(cc ~ rs1419383 + ethnicity, data = imp)
R> pool(fit)
\end{verbatim}
The function \texttt{mice} by default performs five imputations and
therefore five analyses whose results are combined with
\texttt{pool} command. We can also load Bioconductor
(\url{http://www.bioconductor.org/}) package \pkg{RSNPper}
\citep{Carey06} and report metadata (e.g., genome annotation) and
population information based on hapmap, as follows.
\begin{verbatim}
R> library("RSNPper")
R> mysnp <- SNPinfo("rs1419383")
R> popDetails(mysnp)
R> geneInfo("HNF4A")
\end{verbatim}
Markers involved in a multistage design can largely be seen as
monotonic missing by design and can be dealt with similarly. For
large data from a typical genomewide association involving much more
SNPs, we can use database connection mechanisms such as open
database connectivity (ODBC).
\subsection{The EPIC-Norfolk study of obesity}
Our EPIC-Norfolk genomewide association study of obesity serves as a
good example which many principles just outlined apply. The study
was initially designed as a case-control study with the following
definition of cases and controls: cases are those with body mass
index (BMI) $> 30 kg/m^2$ and controls with $20 kg/m^2 \leq
\mbox{BMI} < 25 kg/m^2$. This leads to all obese individuals of the
EPIC-Norfolk cohort (3425), and 3400 controls. Half of the samples
are to be genotyped at stage one, to be followed by the remaining
half at stage two. When adapting this into a case-cohort design, we
wished to know the power/sample size required and this turned to be
straightforward to obtain. Part of calculation in \cite{cz04} and
the required sample size or power in our case can be obtained with
\code{example("ccsize", package = "gap")}. It turned out a
comparable case-cohort design has approximately 2500 controls.
Interestingly, this also corresponds to about 10\% of the size of
our cohort ($n$), which would expect to give stable estimate for a
range of covariates measured in the cohort.
\section{Summary}
This short report shows that the design and implementation of the
\pkg{gap} package allows for a wide range of functions available in
a unified fashion. The package is still under development and can be
seen as a prototype for a future more fully established environment.
As a reviewer pointed out that ``A strength of the `\pkg{gap}'
package is that it aims to be the seed point for a general
compilation of such methods, which means it implements many
different kinds of methods, not just the ones developed by the
author'', the author would like to take this opportunity to invite
comments, suggestions and contributions to consolidate the package.
Because of the broad scope of applications, we focus on genomewide
association as a particular application. Part of Table~\ref{asptdt},
Table~\ref{ccpower}, and the use of multiple imputation in staged
design, have not been seen in the literature. The latter requires
more elaborate development with respect to probability weighting.
Function \texttt{tscc} is a much more transparent implementation
than the original authors' \proglang{C}/\proglang{C++} software. For
power calculation, it is desirable to implement other commonly used
models such as additive, dominant and recessive models. Recently,
\cite{skol07} further examined the design issues associated with
two-stage design. I am aware of related efforts which may be useful,
such as \pkg{pbatR} \citep{hoffmann06}, \pkg{SNPassoc}
\citep{gonzalez07}, \pkg{GenABEL} \citep{ aulchenko07} and
\pkg{snpMatrix} \citep{clayton07}. Among others, the \pkg{RSNPper}
\citep{Carey06} is useful for accessing SNP metadata while
\pkg{biomaRt} \citep{Durinck06} enables a large collection of
biological data to be available in \proglang{R}. The focus here is
largely single point analysis, but multilocus methods such as
haplotype analysis have been implemented in \pkg{gap}. For instance,
the function \texttt{hap.score} is an adaptation of
\texttt{haplo.score} function based on \cite{srt02} which can easily
take haplotype frequency estimates and haplotype assignments from
other source and use them within the generalized linear models
framework.
Although the availability of SNP data is overwhelming, the \pkg{gap}
package was conceived such that it will handle not only SNPs but
multiallelic loci and X-chromosome data. Examples include
\texttt{genecounting} and \texttt{hap}, with the former being
flexible enough to include features such as X-linked data and the
latter being able to handle large number of SNPs. But they are unable
to recode allele labels automatically, so functions \texttt{gc.em}
and \texttt{hap.em} are in \texttt{haplo.em} format and used by a
modified function \texttt{hap.score} in association testing. Owing
to limitation in timing, we have used \proglang{SAS} for many tasks
in analyzing our obesity data \citep{zhao07} which partly
contributes to the delay in consolidating SNP specific analyses in
the package. Nevertheless, it will be of value to adapt some of the
utilities developed in \proglang{SAS}.
After the package was posted to CRAN, other packages have appeared
with somewhat overlapping functions, notably \pkg{powerpkg}
\citep{powerpkg}; a recent version of \pkg{genetics} package
\citep{warnes03} also contains a function \texttt{power.casectrl}.
It would be useful to make some comparisons. Another point relates
to the sequence of development of functions for a package. In
retrospect, it would have been easier if all codes were implemented
in native \proglang{R} language and avoiding foreign language calls,
but the latter may be advantageous in speed given some tuning is
done. Notably, the following functions need further work:
\texttt{bt}, \texttt{kin.morgan}, \texttt{twinan90} \citep{wcn92},
\texttt{gcontrol} \citep{dr99}, \texttt{mtdt} \citep{sham97},
\texttt{s2k} \citep{hirotsu01}. The overall target would involve
better memory management, maintaining numerical precision or
extensions, and use more native \code{.Call} and \proglang{S4}
classes of \proglang{R}. It is likely that existing functions and
data would continue to evolve.
The major driving force to integrate many tools for genetic data
analysis is in line with the observation that statistical genetic
methods or genetic epidemiology is increasingly becoming part of the
general epidemiology with focus on both genetic and nongenetic
factors for diseases, where statistical and computational methods
are more established. On this point, I wish to emphasize the benefit
of shifting to or working on \proglang{R} from my own experience,
which has been a very rewarding one, e.g., \cite{zhao05} and
\cite{zhao06} in relation to package \pkg{kinship}
\citep{therneau03}. I noted a similar but more formal effort has
been launched (\url{http://rgenetics.org/}) but certainly an
informal approach also has its place, as for example in the case of
\pkg{haplo.stats} \citep{sinnwell07}. The best data structures have
yet to be established and collaborative work would be beneficial. I
hope eventually this will be part of a bigger effort to fulfill most
of the requirements foreseen by many e.g., \cite{guo00}.
\section*{Acknowledgments}
The \proglang{R} development team, particularly Prof.~Kurt Hornik,
gave many inputs during the package development. Dr.~Anthony Long
kindly provided me their results similar to Table~\ref{ccpower}
based on an approximation algorithm. Dr.~Jian'an Luan provided the
chromosome 20 data. Dr.~Mike Weale kindly read through the manuscript
and provided useful comments. Comments from anonymous referees and
the editor Achim Zeileis also help to improve the manuscript.
%\bibliographystyle{jss}
\bibliography{jss}
\end{document}