Table 3.
SNP selection P‐value threshold | N SNPs | PHS | PRS | Correlation between PHS and PRS | ||
---|---|---|---|---|---|---|
B [SD] | P‐value | B [SD] | P‐value | |||
5 × 10−8 | 31 | 0.28 [0.04] | 4.3 × 10−13 | 0.28 [0.04] | 2.5 × 10−12 | 0.96 |
10−5 | 80 | 0.26 [0.02] | 2.8 × 10−11 | 0.29 [0.05] | 5.7 × 10−11 | 0.79 |
10−3 | 1460 | 0.13 [0.025] | 3.4 × 10−3 | 0.18 [0.04] | 1.9 × 10−4 | 0.17 |
0.05 | 29998 | 0.07 [0.027] | 0.12 | 0.1 [0.034] | 0.04 | 0.21 |
0.1 | 49247 | 0.06 [0.031] | 0.18 | 0.09 [0.028] | 0.05 | 0.27 |
0.5 | 128952 | 0.08 [0.024] | 0.07 | 0.08 [0.035] | 0.13 | 0.42 |
First column shows the P‐value thresholds for AD associated SNP selection (from an independent IGAP_noGERAD data). Second column represents the number of SNPs that were included to the PRS/PHS score. PHS and PRS effect sizes (mean and SD) and averaged P‐values across 5‐fold cross‐validation subsampling are shown in columns 3–6. The last column shows the average of Pearson's correlation coefficients between PHS and PRS.