166
B. F. Voight
7.3
Uncovering the Biology and Architecture of Complex
Traits
Today, the human genetics field has identified hundreds of reproducible associations
across a range of genetic and pharmacogenetic traits (Hindorff et al. 2009). While
these data represent only a fraction of the underlying genetic contributions, advancements across numerous traits have led to clear insight into the architecture, as well
as clues about the pathophysiology and mechanisms contributing to complex traits.
One can expect additional observations, features, and partial resolution of some of
these questions to emerge as studies continue to amass and specific observations are
taken into functional and experimental systems for further mechanistic proof and
understanding.
7.3.1 Complex Traits Are Differentially Complicated
One clear message is that complex traits broadly share architectural features in
common, but not all traits are uniformly easy to dissect. One can see such features by
comparing the trends in the sample sizes that have led to discoveries across a handful
of traits (Fig. 7.1). Not plotted here are examples in the pharmacogenetics context,
where a few point mutations explain a large fraction of patients with adverse drug
response (SEARCH Collaborative Group et al. 2008; Ge et al. 2009), but we should
be reminded of these cases for their relatively simple architecture. Overall, these
trends indicate a roughly linear relationship between sample size and discovery at
the current stage of analysis for these traits. This implies that, eventually, many
traits will enjoy a similar degree of success in locus discovery efforts, as a function
of samples contributing. Also, these trends indicate different slopes across traits that
imply that each trait has an intrinsic but variable mapping difficulty. The differences
in difficulty could be explained by many factors (e.g., the underlying genetic model,
effect sizes, frequencies of risk variants, the extent of genetic heterogeneity, etc.),
and further dissection of the architecture of traits is a source of active research.
Also contributing to the trait complexity is the degree to which associations are
shared across phenotypes. For example, great success has been met mapping genetic
association for a range of autoimmune disease (AID, e.g., type-1 diabetes (T1D),
inflammatory bowel disease, rheumatoid arthritis). As associations have poured
in, one leading observation is that SNPs with proven association to one trait are
often (and nonrandomly) associated with multiples AIDs (Cotsapas et al. 2011).
This observation is strongly suggestive of shared pathways and biological networks
across diseases, as well as sets of networks that are disease-specific (e.g., the INS
locus for T1D). Metabolic, cardiovascular, and anthropometric networks also appear
to share genetic associations in common, though not nearly the extent as AID:
commonality is more rarely shared among individual SNPs but is instead localized
to discrete genomic locations. As traits share an overlap in genetic causality, one
B. F. Voight
7.3
Uncovering the Biology and Architecture of Complex
Traits
Today, the human genetics field has identified hundreds of reproducible associations
across a range of genetic and pharmacogenetic traits (Hindorff et al. 2009). While
these data represent only a fraction of the underlying genetic contributions, advancements across numerous traits have led to clear insight into the architecture, as well
as clues about the pathophysiology and mechanisms contributing to complex traits.
One can expect additional observations, features, and partial resolution of some of
these questions to emerge as studies continue to amass and specific observations are
taken into functional and experimental systems for further mechanistic proof and
understanding.
7.3.1 Complex Traits Are Differentially Complicated
One clear message is that complex traits broadly share architectural features in
common, but not all traits are uniformly easy to dissect. One can see such features by
comparing the trends in the sample sizes that have led to discoveries across a handful
of traits (Fig. 7.1). Not plotted here are examples in the pharmacogenetics context,
where a few point mutations explain a large fraction of patients with adverse drug
response (SEARCH Collaborative Group et al. 2008; Ge et al. 2009), but we should
be reminded of these cases for their relatively simple architecture. Overall, these
trends indicate a roughly linear relationship between sample size and discovery at
the current stage of analysis for these traits. This implies that, eventually, many
traits will enjoy a similar degree of success in locus discovery efforts, as a function
of samples contributing. Also, these trends indicate different slopes across traits that
imply that each trait has an intrinsic but variable mapping difficulty. The differences
in difficulty could be explained by many factors (e.g., the underlying genetic model,
effect sizes, frequencies of risk variants, the extent of genetic heterogeneity, etc.),
and further dissection of the architecture of traits is a source of active research.
Also contributing to the trait complexity is the degree to which associations are
shared across phenotypes. For example, great success has been met mapping genetic
association for a range of autoimmune disease (AID, e.g., type-1 diabetes (T1D),
inflammatory bowel disease, rheumatoid arthritis). As associations have poured
in, one leading observation is that SNPs with proven association to one trait are
often (and nonrandomly) associated with multiples AIDs (Cotsapas et al. 2011).
This observation is strongly suggestive of shared pathways and biological networks
across diseases, as well as sets of networks that are disease-specific (e.g., the INS
locus for T1D). Metabolic, cardiovascular, and anthropometric networks also appear
to share genetic associations in common, though not nearly the extent as AID:
commonality is more rarely shared among individual SNPs but is instead localized
to discrete genomic locations. As traits share an overlap in genetic causality, one