Table 1.
Outcomes versus coefficients.
Fig 1.
Raw means of sibling standard deviations (before age, sex, mean of trait, and parental genotype controls) by count of minor alleles.
The graph shows that the sibling standard deviation increases with the count of minor allele snps that have effects on variance only, or effects on both the mean and variance of a trait, while stays flat for snp’s that only effect the mean or that not associated with the trait.
Fig 2.
Results of sibling standard deviation method across 1000 replicates.
The figure shows that both in the presence and absence of family-level confounding between the genotype and outcome variable, the method, which examines the effect of an additional minor allele in the sibling pair on the trait’s standard deviation, correctly estimates no variance effects (β = 0) when the outcome is simulated to have mean effects only, and correctly detects variance effects (β ≠ 0) when the outcome is simulated to have variance effects only.
Fig 3.
Manhattan plot of sibling variation in height among FHS 3rd generation sibling pairs.
Results for the pairwise sibling standard deviation in height regressed against the sibling-pair minor count of alleles with controls for sex of sibship, mean age of siblings, age difference of siblings, sibling mean height, parental genotype.
Fig 4.
Manhattan plot of sibling variation in BMI among FHS 3rd generation sibling pairs.
Results for the pairwise sibling standard deviation in BMI regressed against the sibling minor allele count with controls for sex of sibship, mean age of siblings, age difference of siblings, sibling mean BMI, and parental genotype.
Fig 5.
Test for spurious association with variance due to non-linear effects on mean levels.
Mean and standard error for height (inches) and BMI among with two minor alleles is shown separately for homozygotes (one sibling with zero minor alleles and the other sibling with two) and heterozygotes (each sibling has one minor allele), for each genome-wide suggestively significant SNP for the respective trait (A. height; B. BMI). One significant SNP for height (rs8029740) is not depicted because there is only one sibling pair with the 1-1 allele combination and 0 sibling pairs with the 0-2 combination. A two-sample t-test for equality of means, estimated separately for each SNP, revealed no significant differences between the two groups for the top hits for each trait.
Table 2.
Enriched canonical pathways for height and BMI sibling-pair standard deviations in FHS, estimated using i-GSEA4GWAS.
Table 3.
Distribution of third generation siblings included in data by sibship size.