Biomedical Engineering Reference
In-Depth Information
Linkage methods, both parametric and non-parametric, will be outlined next. Although
large-scale association studies are now becoming more prevalent as a means for discover-
ing novel disease genes, and some researchers are questioning what the role of classical
family-based linkage analysis will be in this new era, new linkage findings continue to be
reported and will inform and influence the design and interpretation of future association
studies. Furthermore, active research into combined use of linkage and association infor-
mation in family samples [13] continues to highlight the relevance of linkage analysis for
today's researchers.
Investigators will wish to know what study design is optimal for detecting genes involved
in their disease of interest. It is not always easy to determine a clear, simple answer to this
question. However, this chapter will outline strengths and weaknesses of different methods
and describe some of the implications of particular design choices.
4.2 Methods and approaches
4.2.1 Association methods: unrelated case-control samples
We will first discuss the setting in which the samples are unrelated case and control individ-
uals. The advantages of this study design include the relative ease of ascertaining unrelated
participants compared with families. There are some special issues that require close atten-
tion when using unrelated samples; for example, the cases and controls should be selected
to minimize the possibility of cryptic population structure (population stratification) which
can increase Type I error (false positives). Family-based samples can be more difficult to
recruit and the resulting samples are less efficient in terms of the sample size needed for a
given level of power, compared with a case - control sample [14]. However, family-based
designs are sometimes chosen because they are robust to population stratification.
Owing to the popularity of SNP genotyping for large-scale association mapping projects,
most of our discussion will presume that SNP markers with exactly two alleles (bi-allelic)
are being used. However, we will comment on some instances in which methods discussed
are also appropriate for multi-allelic markers (such as microsatellites).
Protocols for carrying out the analyses discussed will depend in great part on the ana-
lytic software chosen by the investigator. Rather than outlining multiple software-specific
protocols, we discuss the conceptual basis of methods, cite available software packages and
refer interested investigators to the documentation provided with those particular programs.
For illustration, we will also present two general protocols in the context of a large-scale
association study of unrelated case - control samples: a protocol for study design and data
quality control, and a protocol for analysis based on a particular analysis method (logistic
regression).
4.2.1.1 Association in unrelated case-control samples:
study design and data quality
Here we discuss some design issues, and also methods for preprocessing the data before
carrying out statistical tests of association with disease. These include data cleaning checks
and tests for population substructure. Then we will discuss the major statistical methods used
for testing association between genetic markers and disease status in unrelated case -control
samples.
