Additional statistics.
For each 23,574 oligonucleotides represented on the array, we computed a linear regression analysis to test for a correlation between the trait “gonadal fat mass” and each transcript. Similar to our method of calculating linkage, we employed a stepwise linear regression procedure using equations of the form
or
compared to the null model
where β0 is the intercept, and β1, β2, and β3 represent the regression coefficients of their respective terms. As before, the parameter “sex*gene” is only retained if it significantly improves the fit of the model. The p-value threshold for significant correlation is calculated by an F test, which compares the appropriate model (Equation 7 or 8) to the null model (Equation 9). As before, multiple testing was addressed with use of the FDR, ranking the p-values obtained from the above F tests and setting α = 0.01.
Genes correlated with the gonadal fat mass trait generated several significant eQTLs. In order to determine if eQTLs generated by these genes were enriched in any locus or if they were distributed randomly, we compared the distribution of these eQTLs against the distribution of all liver eQTLs in overlapping 6-cM bins across the genome using the Fisher exact test. This test is based on exact probabilities from a specific distribution (hypergeometric distribution). p-Values obtained from this test were corrected for multiple comparisons using a simple Bonferonni correction (given that we performed 772 tests across the genome, 0.05/772). Loci with p < 6.5 × 10−5 by Fisher exact test were considered significantly enriched for eQTLs correlated with gonadal fat mass.