On the modelling of variance components in classical twin studies

Classical twin studies can be used to disentangle the extent to which phenotypic variance of a given complex trait is determined by genetic and environmental variance. The designs widely employ ‘ACDE’ structural equation models where partitioned variances, including that of additive (A) and dominance (D) genetic components, are estimated and where A is taken as reflective of the narrow-sense heritability. Here, it is illustrated in a clear and accessible manner that it is in reality impossible to reliably partition the genetic variance into A and D components using the models which are consequently very open to overestimating the additive genetic variance. This essay should serve as a reminder that classical twin studies can approximate the total (broad-sense) heritability of complex traits, but that leveraging the findings with molecular measurement-based methods is necessary to reliably partition genetic variance components; brief examples based on recent relevant findings is also presented.