. 2020 Mar 20;28(8):1056–1065. doi: 10.1038/s41431-020-0603-2

Table 2.

PRSice-2 polygenic risk score (PRS) prediction results for TwinsUK target sample, for optimal p value threshold.

Measure pair	Threshold	PRS R²	Full R²	Null R²	Coefficient	Standard error	p	p_mixed	No. of SNP	Empirical p
HAID–HAID	0.0056	0.0086	0.1923	0.1836	10,395.2	3121.25	0.0008	0.0012	4266	0.0109
HAID–HD	0.00025	0.0022	0.0748	0.0726	958.2	398.569	0.0162	0.0168	377	0.1461
HAID–PTA	0.00055	0.0074	0.3233	0.3159	26,501.5	8898.04	0.0030	0.0119	712	0.0377
HAID–SNR	0.00055	0.0007	0.1114	0.1107	1824.8	1967.61	0.3539	0.396	712	0.9737
HD–HAID	0.12	0.0113	0.1950	0.1836	21,895.5	5754.99	0.00014	0.00022	46,877	0.00228
HD–HD	0.48	0.0186	0.0911	0.0726	36,885.8	5318.89	4.066E−12	5.23E−11	100,190	<0.0001
HD–PTA	0.37	0.0182	0.3341	0.3159	36,4042	77,141.3	2.786E−06	3.36E−06	88,278	<0.0001
HD–SNR	0.14005	0.0047	0.1153	0.1107	25,480.3	10,639.8	0.0168	0.0281	51,433	0.1122

Measure pair denotes base sample polygenic risk scores constructed from the UK Biobank genome-wide association study on British samples of either wearing a hearing aid (HAID) or having hearing difficulties (HD), respectively, predicting the TwinsUK target phenotypes HAID, HD, pure-tone audiogram (PTA), and signal-in-noise (SNR), respectively. Threshold is the p value threshold for SNP inclusion that maximised variance explained in the respective target phenotype. PRS R² is the variance in the target phenotype explained by the PRS. Full R² is the variance explained by the full-model regression, which includes the PRS as well as the covariates age and sex (no sex covariate for PTA, since there were no males in that sample), while Null R² does not include the PRS, just the covariates. Coefficient is the regression coefficient for the PRS term, with Standard Error and p its corresponding standard error and p value. P_mixed was obtained from running a mixed model accounting for correlation among dizygotic twin pairs. No. of SNP is the number of SNPs included in the PRS and empirical p is the p value obtained from simulation, which corrects for both multiple thresholds tested in order to obtain the optimal threshold and for relatedness in the sample, such as the inclusion of DZ twin pairs.