LJI308 br Materials and Methods br

Materials and Methods
Results
Discussion This variance-component analysis study assessed five genetic and familial environment risk contributions to MDD and SDD using GS:SFHS, a Scottish sample comprising close and distant relatives with genome-wide genotyping data. We showed that the common variant- and pedigree- associated LJI308 and the couple-shared environmental effects are major factors influencing liability to both clinically assessed and self-reported depression among the factors considered. The estimated correlation between SDD and MDD was very high (r=1.00, se=0.20) for common variant-associated genetic effects and lower for the pedigree-associated genetic (r=0.57, se=0.08) and couple-shared environmental effects (r=0.52, se=0.22). The estimates of both common-variant-associated and pedigree-associated genetic effects are much lower for MDD (h2=10%(5%), h2=20%(12%)) than for SDD (h2=22%(6%), h2=50%(15%)), suggesting a difference between clinical and self-reported depression definitions. MDD is diagnosed through clinical questionnaire, whereas SDD reflects both depression status and the participants\' self-awareness, this difference may underlie the higher heritability in SDD. The point estimate of SNP heritability h2 for MDD (h2=10%(5%)) is lower than that from a mega-analysis of 9 cohorts of European ancestry (21%(2%)) (Lee et al., 2013). This may be due to the intrinsic heterogeneity of MDD (Major Depressive Disorder Working Group of the Psychiatric et al., 2013). The pedigree-associated genetic component h2measures the additional genetic effects co-segregating in the pedigree (i.e. those not associated with common genetic variants), such as the effect of rare and structural variants. Using the full model which partitions the narrow sense heritability h2 into the common-variant-associated componenth2and the pedigree-associated component h2,and accounts for multiple familial environmental effects simultaneously, the estimated h2accounted for around one third of h2 in both MDD and SDD. This is lower than previous estimates for other complex traits (excluding depression) that suggest that >50% of h2 is accounted by h2 (Zaitlen et al., 2013; Xia et al., 2016). Here h2accounted for more than two thirds of h2, suggesting an important role for rare and structural variants in both clinical- and self-declared depression. The full model included a number of correlated matrices, potentially impeding model-fitting and their estimation (the discussion for collinearity is Text s3, Tables s4 and s5). A greater discriminating power may however have be achievable with larger sample sizes (Xia et al., 2016). In addition, the power to detect the effect might vary among tested components, therefore a caution should be made when compare the estimates of those components. Stepwise model selections suggested that the two genetic effects and the couple-shared environmental effects were the most significant contributors to risk of depression among the factors considered ( model was selected by both forward and backward selections for SDD and backward selection for MDD), although there is inconsistency in the results between the backward and the forward selections for MDD. This is probably due to the high correlation between ERMfamily and the combined ERMcouple and GRMkin (Xia et al., 2016). Simulation analysis of the backward selection method suggested that although it successfully selected the appropriate model with all major components of simulated phenotype in >80% of cases, in 20% of cases the model either selected the component when the true components were plus or the model selected plus but the true component was (Xia et al., 2016). Given that the pedigree-associated genetic component (K), a component that has been shown to be a major contributor of the genetic effect by the GK and the full model GKFSC in both MDD and SDD, was excluded by backward selection for MDD, it is likely we met the same problem as observed in simulation analysis by Xia et al. and the selected GF model is unlikely to contain all major resources of variation for MDD. As shown in simulation analysis, we expect that a larger sample size may provide sufficient power for a higher accuracy and stability of backward model selection (Xia et al., 2016). In addition, the estimates in the full model for MDD and SDD also suggest that components omitted from model are likely to contribute only small amounts of variance, therefore model was chosen as the best model for both traits in GS:SFHS. The significant contributions from the , and should be further replicated in independent samples.