Three methods for estimating heritability are detailed in the Supplementary Note. The primary approach was twin-based heritability via a REML mixed model, with random additive genetic components of variation, along with shared and individual-specific environmental effects, plus selected covariates as fixed effects. Random terms were assumed mutually independent and normally distributed with mean 0 and variances σa2,σc2,and σe2. This corresponds to a standard ACE model, and assumes DZ twins have an average identity-by-descent proportion of 0.5. 82,83 For each transcript, the twin-based heritability and shared environmental effects were estimated as a^2=σ^a2/(σ^a2+σ^c2+σ^e2) and c^2=σc2/(σ^a2+σ^c2+σ^e2). The ACE model can be fit using either variance-component maximization, constraining â2 and ĉ2 to be non-negative, or using an unconstrained general covariance structure. After establishing that results from the two approaches were highly concordant, we used the unconstrained approach in order to best match the intra-class correlation approach81 used for pathway analysis. Under additive assumptions, â2 is the heritability estimate h2, and P-values are reported for the right tail (positive â2) except where noted. P-values for the X-chromosome were obtained using separate heritability analysis for males
