Biomedical Engineering Reference
In-Depth Information
available; among the best known are the mean test and the proportion test. The former has
been shown to perform best for dominant traits and the latter is best for recessive traits [83].
This general idea of comparing IBD status and phenotypic status also has been extended to
analysis of quantitative traits [84].
VC linkage methods, implemented in programs such as SOLAR [86], ACT [87] and
SEGPATH [88], are well suited to analysis of quantitative traits but may also be applied to
dichotomous or polychotomous traits via a threshold liability model. The typical approach
uses the multivariate normal distribution to model the likelihood of a pedigree, and thus
departure of the data from this assumption of multivariate normality can be a concern [89].
One alternative is to use the multivariate
t
distribution, an option implemented in SOLAR.
Because modern techniques have enabled dense genotyping of markers, non-parametric
sib-pair studies typically generate data at multiple markers across each chromosome and
analyze the data using multipoint methods. The programs MERLIN and Genehunter have
already been mentioned above; these implement both parametric and non-parametric multi-
point linkage analyses, and are now more widely used than the earlier landmark programs
MAPMAKER/SIBS [90] and Neil Risch's ASPEX (http://aspex.sourceforge.net/) for mul-
tipoint sib-pair analysis.
Many of the programs mentioned above have historically required familiarity with UNIX
operating systems and with formats for the input and output of the specific programs. The
recently developed package easyLINKAGE-Plus [91] provides a user-friendly, automated
way to set up and run analyses using any of several popular programs for both para-
metric and non-parametric linkage, and it may be applied to either SNP or microsatellite
markers.
4.2.5 Summary and conclusions
We have described association- and linkage-based approaches to mapping genes influenc-
ing complex traits. Recent successes using genome-wide association mapping designs have
brought attention both to specific genes involved in key biological processes and to this
general approach as a feasible and effective way to identify genes that harbor common
variants involved in common complex diseases. However, a limitation of the genome-wide
association approach is that it is well-powered to detect association only for common alle-
les. Once such a variant is identified, the surrounding gene(s) or region may certainly be
investigated further for additional rare variants that may also influence disease outcomes.
However, to detect a locus at which multiple rare variants and no common alleles cause
the disease, we may need still to rely on classical linkage analysis. In any case, no single
approach should be expected to fit all settings, and the range of strategies discussed here
will remain relevant even as the field of human genetics evolves.
4.3 Troubleshooting
4.3.1 Combining datasets
Sometimes a study will need to combine two datasets that were genotyped and cleaned
separately; for example, two distinct samples that have been genotyped on two overlapping
sets of SNPs. A merged dataset may increase the sample size, the number of SNPs, or both,
