Table 4.
Power (%) of RAML and SKAT methods to detect association of rare variants under seven scenarios for underlying genetic architecture using data simulated for BRCA1 in 4000 cases and 4000 controls
| |
|
Threshold for significance |
|||||
|---|---|---|---|---|---|---|---|
| |
|
P < 0.001 |
P < 0.01 |
P < 0.05 |
|||
| Scenario* | Proportion of variants associated | RAML | SKAT-O | RAML | SKAT-O | RAML | SKAT-O |
| 1 |
0.05 |
47.5 |
31.5 |
61.0 |
44.0 |
73.5 |
58.5 |
| 2 |
0.05 |
39.5 |
24.0 |
53.5 |
36.0 |
64.5 |
49.0 |
| 3 |
0.05 |
43.0 |
29.5 |
61.5 |
43.0 |
70.5 |
55.0 |
| 4 |
0.05 |
33.5 |
22.5 |
49.5 |
29.0 |
59.5 |
42.5 |
| 5 |
0.05 |
39.5 |
25.5 |
57.0 |
39.5 |
62.0 |
51.5 |
| 6 |
0.05 |
33.5 |
21.5 |
43.0 |
29.5 |
57.0 |
41.5 |
| 7 |
0.05 |
37.5 |
30.5 |
51.0 |
38.0 |
56.5 |
45.5 |
| 1 |
0.10 |
40.0 |
34.5 |
58.0 |
50.0 |
76.0 |
66.0 |
| 2 |
0.10 |
36.0 |
29.0 |
50.5 |
42.5 |
65.5 |
58.0 |
| 3 |
0.10 |
32.5 |
28.5 |
53.0 |
42.5 |
67.5 |
58.0 |
| 4 |
0.10 |
31.0 |
22.0 |
45.5 |
35.5 |
59.0 |
48.5 |
| 5 |
0.10 |
33.5 |
25.0 |
55.0 |
39.5 |
71.0 |
59.5 |
| 6 |
0.10 |
32.5 |
21.5 |
48.5 |
31.5 |
62.0 |
48.5 |
| 7 |
0.10 |
31.5 |
23.0 |
44.0 |
35.0 |
60.5 |
46.5 |
| 1 |
0.20 |
47.5 |
50.0 |
71.0 |
67.5 |
83.5 |
81.0 |
| 2 |
0.20 |
26.5 |
34.5 |
49.0 |
46.0 |
69.5 |
68.0 |
| 3 |
0.20 |
41.0 |
40.5 |
59.5 |
52.5 |
74.0 |
67.0 |
| 4 |
0.20 |
26.0 |
26.0 |
44.5 |
40.0 |
67.5 |
55.5 |
| 5 |
0.20 |
35.0 |
26.5 |
51.5 |
44.5 |
70.0 |
64.5 |
| 6 |
0.20 |
27.5 |
24.5 |
45.5 |
37.0 |
62.5 |
55.0 |
| 7 | 0.20 | 23.5 | 19.0 | 46.0 | 35.5 | 63.0 | 59.5 |
Method with greatest power emboldened.
* See text for description of genetic architecture for each scenario.
RAML Rare admixture maximum likelihood, SKAT-O sequence kernel association test, T1 fixed threshold test 1 per cent MAF, T5 fixed threshold test 5 per cent MAF, WST weighted sum test, VTT variable threshold test, EREC estimated regression coefficient test.