A Robust Rerank Approach for Feature Selection and Its Application to Pooling-Based GWA Studies
Table 1
Association analysis of the 16 overlapping SNPs in both stages of bipolar study.
SNP
Chr
Pooling (Stage 1)
Individual genotyping (Stage 2)
SAM
CA
CN
CA
CN
Model
OR
value
SNP 1
3
7
189
5
0.17
0.36
0.21
0.35
DOM
0.5
0.00002*
SNP 2
3
2
363
15
0.23
0.43
0.22
0.33
DOM
0.6
0.00018*
SNP 3
6
28
873
152
0.53
0.38
0.45
0.33
ALL
1.6
0.00085*
SNP 4
4
36
124
58
0.31
0.44
0.33
0.44
ALL
0.7
0.00142*
SNP 5
21
13
531
64
0.58
0.41
0.55
0.43
ADD
1.5
0.00294*
SNP 6
6
41
1308
346
0.41
0.28
0.34
0.25
ADD
1.5
0.00651*
SNP 7
16
10
110
18
0.33
0.49
0.34
0.45
ALL
0.7
0.00725*
SNP 8
13
48
596
126
0.56
0.42
0.52
0.40
ALL
1.4
0.00785*
SNP 9
2
101
7484
2795
0.50
0.40
0.48
0.41
ADD
1.4
0.00960*
SNP 10
4
5
19
1
0.27
0.46
0.27
0.36
ALL
0.7
0.00971*
SNP 11
14
40
4678
1025
0.36
0.49
0.42
0.53
DOM
0.7
0.01275*
SNP 12
18
20
393
47
0.49
0.33
0.47
0.36
ALL
1.4
0.01604*
SNP 13
3
6
694
44
0.33
0.53
0.39
0.49
DOM
0.7
0.01631*
SNP 14
7
8
36
2
0.31
0.47
0.37
0.45
ADD
0.8
0.06830
SNP 15
5
21
762
57
0.37
0.56
0.35
0.42
DOM
0.8
0.09806
SNP 16
7
22
210
21
0.35
0.53
0.31
0.38
ALL
0.8
0.13530
Chr: chromosome; : rank based on rerank approach; : rank based on -test; SAM: rank based on SAM with ; CA: MAF of cases; CN: MAF of controls; Model: genetic model with the most significant result among allelic (ALL), dominant (DOM), recessive (REC), and additive (ADD) genetic models; OR (odds ratio) and value: estimated by the simple logistic regression under the genetic model with the most significant result (* value 0.05).