7 What Have We Learned from GWAS?
173
7.4.1 Custom Genotyping Arrays Technologies for Genetic Studies
The central findings from GWAS, where hundreds of associations to multiple (and
related) traits have been identified, are two central questions, among several others.
First, can further genetic investigation identify additional associated regions with
these traits? And second, can detailed genetic investigation localize causal variants
where an established hit has been identified? Addressing these two questions, however, requires assaying hundreds of thousands of genetic variants from thousands of
individuals, which is an expensive, labor-intensive, and time-consuming process. To
make this process cost-efficient and streamlined in terms of producing an analysis,
2010 saw an effort to develop custom-array genotyping technologies, which are
built directly from the information from leading association studies efforts and have
been initiated. These include, among potentially others in development, the IBCchip, the Metabochip for cardiovascular, metabolic, and anthropometric traits, and
the ImmunoChip for AIDS (Keating et al. 2008; Cortes and Brown 2011; Voight et
al. 2012). In addition, a custom array developed to comprehensively genotype lowfrequency and rare variation based on discoveries from exome-sequencing projects
for the coding genome, the Exomechip, has also been developed. These technologies
offer data in a well-powered, second stage, to facilitate replication and fine-mapping
genetic studies, as well as machine-learning or pathway/module-based methods to
identify networks related to disease and to characterize their architecture. The first
set of studies using these technologies are starting now to be published (Trynka
et al. 2011, Morris et al. 2012; CARDIoGRAM Consortium 2013; Global Lipids
Genetics Consortium 2013), and one should expect a surge of new locus discoveries
and statistical methods that utilize this approach to data collection and technology.
7.4.2 Analysis of Multiple Phenotypic Measurements
and Outcomes
The past 5 years of study has focused a large fraction of effort toward the analysis
of individual phenotypes one at a time. An observation emerging from such studies
is that multiple traits overlap in association, either at SNPs or at discrete physical
locations of the genome. This is perhaps the most notable for autoimmune disease
(Cotsapas et al. 2011), though several instances for metabolic (as previously discussed) and psychiatric diseases have also been observed. This set of observations
clearly justify more expansive genetic studies which analyze multiple phenotypes
simultaneously either using summary data or jointly in multivariate regression
models. These approaches can be expected to uncover genetic loci with compelling
associations that have not yet been captured by existing studies due to a lack of
power.
One hypothesis along this line of inquiry posits that regions with associations
to multiple traits are uniquely positioned to help identify causal genes, tissues,
and mechanisms of action contributing to them. This hypothesis relies on the
173
7.4.1 Custom Genotyping Arrays Technologies for Genetic Studies
The central findings from GWAS, where hundreds of associations to multiple (and
related) traits have been identified, are two central questions, among several others.
First, can further genetic investigation identify additional associated regions with
these traits? And second, can detailed genetic investigation localize causal variants
where an established hit has been identified? Addressing these two questions, however, requires assaying hundreds of thousands of genetic variants from thousands of
individuals, which is an expensive, labor-intensive, and time-consuming process. To
make this process cost-efficient and streamlined in terms of producing an analysis,
2010 saw an effort to develop custom-array genotyping technologies, which are
built directly from the information from leading association studies efforts and have
been initiated. These include, among potentially others in development, the IBCchip, the Metabochip for cardiovascular, metabolic, and anthropometric traits, and
the ImmunoChip for AIDS (Keating et al. 2008; Cortes and Brown 2011; Voight et
al. 2012). In addition, a custom array developed to comprehensively genotype lowfrequency and rare variation based on discoveries from exome-sequencing projects
for the coding genome, the Exomechip, has also been developed. These technologies
offer data in a well-powered, second stage, to facilitate replication and fine-mapping
genetic studies, as well as machine-learning or pathway/module-based methods to
identify networks related to disease and to characterize their architecture. The first
set of studies using these technologies are starting now to be published (Trynka
et al. 2011, Morris et al. 2012; CARDIoGRAM Consortium 2013; Global Lipids
Genetics Consortium 2013), and one should expect a surge of new locus discoveries
and statistical methods that utilize this approach to data collection and technology.
7.4.2 Analysis of Multiple Phenotypic Measurements
and Outcomes
The past 5 years of study has focused a large fraction of effort toward the analysis
of individual phenotypes one at a time. An observation emerging from such studies
is that multiple traits overlap in association, either at SNPs or at discrete physical
locations of the genome. This is perhaps the most notable for autoimmune disease
(Cotsapas et al. 2011), though several instances for metabolic (as previously discussed) and psychiatric diseases have also been observed. This set of observations
clearly justify more expansive genetic studies which analyze multiple phenotypes
simultaneously either using summary data or jointly in multivariate regression
models. These approaches can be expected to uncover genetic loci with compelling
associations that have not yet been captured by existing studies due to a lack of
power.
One hypothesis along this line of inquiry posits that regions with associations
to multiple traits are uniquely positioned to help identify causal genes, tissues,
and mechanisms of action contributing to them. This hypothesis relies on the