We next evaluated what fraction of the Lisboa DEG showed evidence for differential expression in our study in two ways. First, we fitted a beta-uniform mixture model as implemented in function “fitBumModel” in R library BioNet to the p-values from our study for the overlapping genes. For this model, the goal is to estimate the fraction of genes in our study inferred to have signal, given that they were DEG in the Lisboa study; i.e., what fraction of observations were drawn from the Beta distribution (Fig. S5). This estimate was 1.0, suggesting that all DEG from the Lisboa study carry signal in our data. In order to further validate the similarity between study results, we next used the function “fdrThreshold” (BioNet) to produce a p-value cutoff that corresponds to an FDR of 0.20 for both regions in our dataset. Test statistics for 310 and 71 genes fell below this FDR threshold for caudate and nucleus accumbens, respectively. Of these, 263 and 64 were congruent in the direction of their log fold change across studies (Fig. S5). Using the set of
