A general method for combining different family-based rare-variant tests of association to improve power and robustness of a wide range of genetic architectures

Table 1 Type I error rate for 10 genes selected to be representative of all 30 genes explored

Gene	\( {Q}_2 \)	\( {Q}_{2^{*}} \)	\( {Q}_1 \)	\( {Q}_{1^{*}} \)	\( {Q}_4 \)	\( {Q}_{4^{*}} \)	\( {Q}_3 \)	\( {Q}_{3^{*}} \)	Combined
High power
ARF4	0.055	0.05	0.015	0.03	0.065	0.04	0.04	0.025	0.03
DNASE1L3	0.055	0.085	0.07	0.07	0.08	0.075	0.075	0.07	0.055
MAP4	0.06	0.05	0.055	0.09	0.085	0.09	0.055	0.06	0.07
SCAP	0.09	0.135	0.025	0.055	0.08	0.18	0.055	0.075	0.055
Low power
CXCR6	0.07	0.045	0.115	0.065	0.075	0.065	0.065	0.045	0.04
PAK2	0.04	0.095	0.045	0.055	0.04	0.08	0.035	0.055	0.035
PTPLB	0.005	0.025	0.055	0.015	0.01	0.045	0.01	0.035	0.03
RAD18	0.045	0.045	0.025	0.06	0.045	0.035	0.04	0.04	0.03
Other representative genes
PDCD6IP	0.05	0.035	0.055	0.045	0.055	0.055	0.05	0.045	0.035
ZBTB38	0.065	0.035	0.045	0.02	0.06	0.055	0.055	0.04	0.03
Mean Across All 30 Genes (SD)	0.067 (0.03)	0.074 (0.05)	0.048 (0.02)	0.059 (0.05)	0.07 (0.03)	0.084 (0.05)	0.058 (0.03)	0.065 (0.05)	0.058 (0.04)

Type I error rates are reported for all 9 tests across 10 different genes: the 4 genes with highest overall power, the 4 genes with lowest overall power, and 2 genes deemed representative of the remaining genes explored. The mean type I error rate for each test statistic across all 30 genes is reported at the bottom of the table. Averaged across all 30 genes, the type I error rate was nominally conserved at a 0.05 level

ISSN: 1753-6561