Skip to main content

Table 2 Comparison of performance of DASSO_MB, BEAM and SVM algorithms

From: A Markov blanket-based method for detecting causal SNPs in GWAS

 

MAF

Model 1 ( r 2 =0.7)

0.05

0.1

0.2

0.5

DASSO-MB

0(0)

0(0)

0(0)

32(0.16)

BEAM

0(0)

0(0)

0(0)

22(0.05)

SVM

1(3)

1(3)

0(0)

33(0.7)

 

MAF

Model 1 ( r 2 =1)

0.05

0.1

0.2

0.5

DASSO-MB

0(0)

0(0)

0(0)

46(0.11)

BEAM

0(0)

0(0)

0(0)

36(0.07)

SVM

0(0)

0(0)

1(2)

43(0.76)

 

MAF

Model 2 ( r 2 =0.7)

0.05

0.1

0.2

0.5

DASSO-MB

0(0)

8(0)

26(0.12)

18(0)

BEAM

0(0)

2(0)

10(0.3)

9(0.11)

SVM

0(0)

2(1.5)

14(0.93)

21(0.8)

 

MAF

Model 2 ( r 2 =1)

0.05

0.1

0.2

0.5

DASSO-MB

10(0)

22(0.05)

42(0.05)

33(0.03)

BEAM

8(0.13)

7(0)

17(0.24)

27(0.11)

SVM

1(2)

3(0.67)

22(1.18)

33(0.94)

 

MAF

Model 3 ( r 2 =0.7)

0.05

0.1

0.2

0.5

DASSO-MB

24(0.04)

44(0.14)

47(0.02)

11(0.09)

BEAM

21(0.14)

24(0)

32(0.09)

11(0.09)

SVM

1(1)

6(1.83)

29(0.83)

29(0.83)

 

MAF

Model 3 ( r 2 =1)

0.05

0.1

0.2

0.5

DASSO-MB

34(0.03)

50(0.08)

49(0.04)

31(0.06)

BEAM

33(0.03)

47(0.04)

43(0.09)

31(0.1)

SVM

5(1.6)

23(1.52)

42(0.64)

38(0.55)

  1. We show the number of datasets in which two disease-associated markers can be identified with no more than two false positives. The average number of false positives is in the parentheses.