Table 1.
Ten height genes implicated by gene-based testing. These genes meet our three criteria for statistical significance: (1) gene-based P<5×10-7, (2) the gene does not include variants with P<2×10-7, and (3) the gene-based P-value is at least two orders of magnitude smaller than the P-value for the most significant variant within the gene. For each gene, we provide P-values for the four different gene-based tests applied. P-values in bold are the most significant results for a given gene.
| Gene | Discovery gene-based P-value | Validation P-value1 | Combined P-value1 | Conditional P-value2 | Note3 | |||
|---|---|---|---|---|---|---|---|---|
| SKAT-broad | VT-broad | SKAT-strict | VT-strict | |||||
| OSGIN1 | 4.3×10-11 | 4.5×10-5 | 0.19 | 0.18 | 0.048 | 2.6×10-12 | 7.7×10-11 | Known locus. No predicted causal genes. |
| CRISPLD1 | 2.2×10-7 | 6.7×10-11 | 8.5×10-6 | 8.9×10-7 | 0.50 | 1.2×10-12 | NA | Known locus, sentinel GWAS SNP not tested on ExomeChip. CRISPLD1 was predicted to be causal. |
| CSAD | 2.3×10-8 | 2.4×10-9 | 0.83 | 0.59 | 0.54 | 2.0×10-9 | NA | New locus. |
| SNED1 | 1.9×10-5 | 4.3×10-9 | NA | NA | 0.083 | 4.5×10-10 | 1.4×10-9 | Known locus. SNED1 was not predicted to be causal. |
| G6PC | 1.3×10-5 | 3.6×10-8 | 5.5×10-6 | 1.3×10-6 | 0.24 | 5.2×10-8 | 3.9×10-8 | Known locus, G6PC was not predicted to be causal. G6PC is mutated in glycogen storage disease Ia. |
| NOX4 | 5.1×10-6 | 1.4×10-7 | NA | NA | 0.013 | 5.5×10-9 | NA | New locus. |
| UGGT2 | 3.0×10-5 | 2.6×10-7 | 2.3×10-5 | 4.8×10-7 | 0.64 | 3.4×10-7 | NA | New locus. |
| FLNB | 2.2×10-6 | 5.1×10-4 | 2.4×10-9 | 3.2×10-6 | 0.016 | 8.6×10-11 | 3.6×10-9 | Known locus. FLNB was predicted to be causal. FLNB is mutated in atelosteogenesis type I. |
| B4GALNT3 | 2.4×10-5 | 1.9×10-5 | 1.8×10-5 | 3.1×10-7 | 0.79 | 4.3×10-7 | 7.7×10-7 | Known locus. B4GALNT3 was predicted to be causal. |
| CCDC3 | 6.3×10-4 | 6.3×10-6 | 3.0×10-7 | 5.4×10-9 | 0.080 | 1.2×10-9 | 1.6×10-9 | Known locus. CCDC3 was predicted to be causal. |
Validation (N=59,804) and combined results using the same test and (when possible) variants.
When the gene is located in a locus identified by our single-variant analysis (1 Mb window), we conditioned the gene-based association result on genotypes at the single variant(s).
If the gene falls within a known GWAS height locus, we mention if it was predicted to be causal using bioinformatic tools (ref. 3). NA, not applicable.