seen in the rate of de novo synonymous variants (RR 0.91, 95% CI 0.70–1.17, p = 0.8) nor in the rate of in-frame indels (RR 0.45, 95% CI 0.019–3.48, p = 0.9). A one-sided binomial exact test, which is typically used in WES studies to assess the significance of observed burden differences in cases versus controls (e.g., Iossifov et al., 2014; Sanders et al., 2012; Willsey et al., 2013), produced consistent results (Figure S2). Indeed, these results more strongly support the association of de novo variants with TD. However, as the distribution of callable base pairs per sample varied across the cohorts due to differences in experimental design (e.g., library capture protocol or sequencing coverage; Table 1; Figure S3), we felt the rate ratio test would provide a more accurate estimate of the significance of true variant burden. This approach compares the number of variants while also controlling for the per sample differences in the number of base pairs with sufficient coverage and quality for de novo detection.
