missing heritability
Hemani G, Knott S & Haley C 2013 An evolutionary perspective on epistasis and the missing heritability. PLoS Genet 9:e1003295.
- one reason that the problem of the "missing heritability" arises is because the additive genetic variation that is estimated to be contributing to the variance of a trait will most likely be an artefact of the non-additive variance that can be maintained over evolutionary time
- even a small reduction in linkage disequilibrium between causal variants and observed SNPs rapidly erodes estimates of epistatic variance, leading to an inflation in the perceived importance of additive effects
- the perception of independent additive effects comprising the majority of the genetic architecture of complex traits is biased upwards
- the search for causal variants in complex traits under selection is potentially underpowered by parameterising for additive effects alone
- the detection of causal variants through genome-wide association studies may be improved by searching for epistatic effects explicitly
- epistatic interactions can allow deleterious mutations to persist under selection
- these interactions can abate the depletion of additive genetic variation
- a much larger element of non-additive genetic variance is maintained
- the heritability estimated from family studies could be a mixture of both additive and non-additive components
- searching directly for epistatic effects greatly improves the discovery of variants under selection, despite the multiple testing penalty being much larger
- common practices in genome-wide association studies could lead to both an ascertainment bias in detecting additive effects and a confirmation bias in perceiving that most of the genetic variance is additive
- traits under selection often evolve more slowly than expected
- contrary to expectation, genetic variation is maintained under selection
- this problem is known as 'stasis'
- it is particularly evident in fitness-related traits where the genetic variation tends to be highest
- yet there is commonly no observed response to selection at all
- by definition epistasis will form a part of the 'unknown heritability'
- epistatic interactions could also contribute to h2 estimates
- this could arise through two possible mechanisms
- firstly by generating real additive variation as marginal effects from higher order genetic interactions
- secondly by creating a statistical illusion of additive variance through confounding between non-additive and common environment effects in twin study based estimates
- at the intermediate, within-population level there is a distinct lack of evidence for any widespread importance of epistasis arising from natural variation
- most genetic variation appears to be additive
- for many of the patterns of epistasis that we assayed, deleterious effects can be maintained at intermediate frequencies over long evolutionary time periods
- a small amount of additive variation is maintained by epistasis but most genetic variation is non-additive
- there is a strong bias in GWA studies that lead to an overestimation of additive effects at QTLs
- perhaps counterintuitively, the most powerful way to uncover additive variation under selection is to parameterise the search to include epistatic effects using dense genotype information
- because additive variance decays linearly with LD [46], at low LD they remain detectable leading to an ascertainment bias for additive vs non-additive effects
- even in the best case scenario, where G-P maps were generated to maximise additive variance, total genetic variance was mostly composed of non-additive components
- this finding is in disagreement with a recent study [28], which showed that for various two-locus epistatic models, the deterministic partitions of genetic variance calculated across different frequency distributions were largely comprised of the additive component
- those allele frequencies at which additive variance is high (a large proportion of the frequency spectrum), are evolutionarily unstable
- should epistatic variants be affecting fitness traits then the majority of the variance will be non-additive