Table 1.
Demographics of NIA‐LOAD FBS and EFIGA cohort.
| NIALOAD | |
|---|---|
| Total samples | 3676 |
| Number of cases | 1748 |
| Number of controls | 1928 |
| Number of unrelated individuals | 846 |
| Number of families | 666 |
| Mean number of individuals per family | 4.25 ± 2.67 |
| Average age of cases | 74.66 ± 7.87 |
| Average age of controls | 77.37 ± 7.95 |
| EFIGA | |||
|---|---|---|---|
| Full dataset | Training | Testing | |
| Total samples | 8512 | 5110 | 3402 |
| Number of cases | 3317 | 1986 | 1331 |
| Number of controls | 5195 | 3124 | 2071 |
| Number of unrelated individuals | 6474 | 4011 | 2751 |
| Number of families 1 | 464 (100%) | 307 (66%) | 211 |
| −45% | |||
| Mean number of individuals per family | 4.39 | 3.58 | 3.09 |
| Average age of cases | 76.53 | 76.42 | 76.68 |
| Average age of controls | 75.5 | 75.55 | 75.42 |
To estimate genome‐wide effect‐sizes of variants in a Caribbean Hispanic population, we randomly divided the full dataset into training and testing sets (preserving the ratio of affected and unaffected individuals). Variant effect‐size estimates for AD‐association were computed in the training dataset across the genome. PRS was computed in the testing dataset using the effect sizes from the training dataset. Penetrance and recurrence risk of the PRS was subsequently computed in the testing dataset.
1
Individuals were randomly split between training and testing datasets, which resulted in some families having members in both training and testing datasets.