Most binary phenotype associations were supported by OMIM or were annotated as pathogenic/likely pathogenic in ClinVar (88.6%), indicating that we robustly captured high-confidence signals (Supplementary Table 10). However, we also identified rare variant associations with phenotypes beyond those reported in OMIM (Supplementary Table 10). For example, 12.1% of the European cohort carried at least one of the 373 distinct filaggrin (FLG) PTVs identified. These individuals had a significantly higher risk of well-known associations, including dermatitis (P = 5.1 × 10−95; OR: 1.96 (95% CI: 1.84–2.08)) and asthma (P = 3.1 × 10−32; OR: 1.24 (95% CI: 1.19–1.28))36, but were also at risk of under-recognized associations, such as melanoma (P = 4.7 × 10−13; OR: 1.21 (95% CI: 1.15–1.27))37 and basal cell carcinoma (P = 9.9 × 10−10; OR: 1.19 (95% CI: 1.12–1.25))38. Concomitant increases in the levels of vitamin D (P = 2.3 × 10−131; β: 0.15 (95% CI: 0.14–0.16))39 suggest that the increased risk of skin cancer may be attributable to increased sensitivity to ultraviolet B radiation. This interrogation offers one example of how this phenome-wide resource can uncover a wide spectrum of phenotypes associated with rare variation in any protein-coding gene.
