Fig. 6: Regional nonsense constraint highlights clinically impactful de novo variants and novel disease gene candidates.

a Odds ratio for receiving a diagnosis (‘case solved’) for 12,303 coding de novo variants in 13,908 rare disease trios in the 100,000 Genomes Project. Error bars represent 95% confidence intervals. Asterisks indicate statistically significant values (two-sided Z test on log-transformed values with Bonferroni correction for 14 tests at alpha=0.05, p < 0.0036). b–e Relative enrichment of coding de novo variants in constrained versus unconstrained regions for 50,825 variants in 32,260 rare disease trios. Relative enrichment is shown for all genes (b), monogenic disease genes associated with autosomal dominant phenotypes in OMIM (c), monogenic disease genes associated with autosomal recessive phenotypes in OMIM (d) and genes with no monogenic disease association in OMIM (e). Error bars show bootstrap 95% confidence intervals. Asterisks show statistically significant values (two-sided percentile bootstrap with Bonferroni correction for 28 tests at alpha=0.05, p < 0.0018). The number of canonical transcripts with a constrained NMD-target region (f) or NMD-escape region (g) harbouring one or more de novo nonsense or frameshift variants in 32,260 rare disease trios. FS = frameshift variant. Exact P values for a, b, c, d, and e are provided in the Source Data file.