fisherian geometry - senecio7の日記

Remington DL 2015 Alleles versus mutations: Understanding the evolution of genetic architecture requires a molecular perspective on allelic origins. Evolution 69:3025-3038.

perspectives on the role of large-effect quantitative trait loci (QTL) in the evolution of complex traits have shifted back and forth over the past few decades
much of the confusion results from a failure to distinguish mutational and allelic effects
a limitation of using the Fisherian model of adaptive evolution as the lens through which the evolution of adaptive variation is examined

inadequate models relating molecular variation to complex phenotypes are severely hampering our understanding of these topics

perspectives on the infinitesimal model began to shift in the 1980s
early QTL studies, generally in artificial populations, almost universally identified chromosomal regions with detectable and often large effects on traits
the resulting quantitative genetic framework, what we might term the "QTL perspective," was still inherently polygenic but not strictly infinitesimal, consistent with Robertson's (1967) model of an exponential distribution of allelic effect sizes

more recent research, however, has challenged the QTL perspective on many fronts
fractionation of large-effect QTL regions into multiple smaller QTL
cryptic effects of QTL that are epistatic or linked to QTL with opposite effects
the rise and fall of the QTL perspective came full circle with a paper by Rockman (2012), who pronounced the entire QTL perspective largely irrelevant to evolutionary biology
Travisano and Shaw (2013) argued further that studies of proximate molecular mechanisms have shed little light on ultimate evolutionary processes
reaffirming the dichotomy between functional and evolutionary biology articulated by Mayr
it is clear that we have entered a new micromutationist era

both the rise and fall of the QTL perspective are based in part on the limitations of premolecular models that conflate beneficial mutations with allelic variation
mutational effects and allelic effects (see Box 1) are not equivalent
different selective regimes operating within and between populations can give rise to very different patterns of complex trait architecture
small-effect mutations and intragenic recombination indeed frequently give rise to adaptively important large-effect alleles under relevant evolutionary scenarios
I will explain why prevailing GWAS approaches consequently do not actually estimate effects of allelic variation
requiring reexamination of GWAS-based conclusions about genetic architecture in both evolutionary biology and biomedical settings
a more suitable framework for understanding the origin of alleles could refocus our attention on more productive paths forward in understanding the genetics of adaptation

none of what I have to say is terribly novel
the issue is one of synthesis
the relevant concepts are largely in place but are not being brought together in a framework that adequately informs our search for answers to some of the big questions in evolutionary biology

Orr's model of sequential fixation of mutations was essentially the same as Fisher's
in essence, the twin poles around which the theoretical debate about the size of mutational effects revolved, represented by Orr's and Fisher's models, remained premolecular
a limitation Orr himself has been careful to point out

a "one QTL-one SNP" paradigm was the foundation for the ensuing development of association mapping strategies from the beginning
in the extensive debates about the nature of the "missing heritability" in GWAS, the idea that alleles differ from one another at a single causative SNP is seldom specifically questioned
the question we must answer, then, is whether the Fisherian allelic perspective and its genome-era counterpart, the one QTL-one SNP paradigm, are realistic

these points highlight the limitations of the Fisher–Orr geometric models as a framework for understanding the nature of adaptive evolution
the notion of sequential fixation of mutations ignores the role of standing allelic variation in adaptation
both the phenotypic and fitness effects of a novel mutation will depend on the mean phenotype of the allele in which a mutation occurs, not to mention allele frequencies at other loci
steps in adaptation are likely to include not only advantageous mutations, but also intragenic recombination events that generate novel higher fitness alleles

the incremental effect of a mutation at the time it first occurred might be substantially different than the current net contribution of the resulting polymorphism to the effects of the alleles in which it is found
we should expect the sizes of mutational effects and allelic effects to follow different distributions, with allelic effects being much larger in many cases
depending on the degree to which individual polymorphisms within a gene interact non-additively, the very concept of mutational effects may be ambiguous in some cases

most of the studies reviewed by Mackay and colleagues (Flint and Mackay 2009; Mackay et al. 2009), which suggest that large-effect loci are atypical, were designed to investigate genetic variation within more-or-less panmictic populations
adaptively important traits in such populations are likely to be under some form of direct or indirect stabilizing selection, which would tend to reduce the frequency of large-effect alleles
individuals carrying them would be more likely to deviate from optimal phenotypes

as explained above, the foundational papers for association mapping were explicitly based on a one QTL-one SNP model
the prevailing approaches used in association mapping studies growing at least implicitly out of this foundation are particularly unsuited to detecting effects at the level of functional alleles

incomplete LD between sampled SNPs and causal polymorphisms is recognized as one of the factors underlying the "missing heritability" in association studies
with the remainder attributed to loci with effects too small to detect
and possibly to inflated estimates of overall additive genetic variance due to epistasis
association studies provide little power to separate the contributions of effect size and allele frequency of causative SNPs to the genetic variance the SNPs explain
the individual causative SNPs in complex alleles may have small effects, segregate at intermediate frequencies, and occur in various combinations

the limitations of the one QTL-one SNP paradigm extend far beyond evolutionary biology

including population structure and kinship in structured association studies can substantially factor out spurious genotype–phenotype associations
if molecular differentiation between populations is strongly correlated with the phenotypic differences between them, the effects of loci underlying between-population differences will also be factored out
one of the first studies to use a structured association approach identified strong associations between sequence variants in Dwarf8 and flowering time variation in maize
this association is largely spurious, and possibly results from selection on multiple traits that are affected by other genes near Dwarf8

my arguments may not satisfy evolutionary biologists who see little relevance of genotype-to-phenotype studies for addressing evolutionary questions in the first place
genotype-to-phenotype studies cannot substitute for direct research on evolutionary processes
studies of molecular detail have added a great deal of richness to evolutionary research by providing insights on how genotypic variation affects evolutionary processes