Quantitative reverse transcription–PCR analyses were conducted in R version 3.0.1 software (R Foundation) using linear modeling, treating log2-transformed expression levels of each transcript as the outcome. Statistical models investigating age-related expression patterns in control participants covaried for RNA integrity number (RIN), sex, postmortem interval (PMI), and race. Subsequent models exploring the influence of these biological variables on the expression of ZNF804A were adjusted for age as a B-spline with knots at 1, 10, 20, and 50 with an offset at birth to capture nonlinear patterns in expression. Statistical models investigating diagnosis-related expression patterns in adults covaried for sex, race, pH, PMI, and RIN; diagnosis-related expression fold changes were relative to the control group. Statistical models investigating genotype-by-diagnosis interactions included main-effect regression terms for genotype and diagnosis as well as an interaction term between the 2 variables to capture genotype effects within each diagnosis group. The statistical models for the genotype effect in fetal samples adjusted only for RIN and race, as the PMI of these samples was short (mean, 2.6 hours) and many samples were missing information on pH. All
