Table 4

Number of times interacting alleles were observed as highest scoring model (rank = 1) or second (rank = 2) using PIA 2.0a for 2-SNP interactions using imbalanced simulated data sets



Scoring Function
Model Numberb
 Case:Control Ratio
 Rank
 Sensitivity +Specificity
 PPV+NPV
 Risk Ratio
 Odds Ratio
 Gini Index
 Probability Difference
 Total (Overall)c
55
 1:2
 1
 72
 72
 69
 72
 81
 74
 82


2
 6
 7
 7
 7
 9
 8
 6

1:4
 1
 56
 50
 50
 50
 61
 63
 67


2
 9
 13
 10
 13
 11
 11
 7
56
 1:2
 1
 69
 67
 59
 67
 83
 73
 79


2
 8
 10
 10
 10
 11
 9
 10

1:4
 1
 45
 43
 38
 43
 62
 52
 62


2
 10
 11
 12
 11
 15
 12
 11
57
 1:2
 1
 63
 62
 45
 62
 82
 69
 76


2
 6
 7
 12
 7
 7
 14
 8

1:4
 1
 51
 49
 34
 49
 67
 53
 66


2
 10
 8
 16
 8
 10
 14
 9
58
 1:2
 1
 87
 86
 79
 86
 96
 92
 95


2
 7
 5
 8
 5
 4
 4
 4

1:4
 1
 73
 70
 66
 69
 84
 79
 84


2
 8
 8
 15
 9
 10
 8
 7
59
 1:2
 1
 64
 61
 62
 61
 72
 72
 70


2
 6
 9
 12
 9
 7
 7
 10

1:4
 1
 54
 50
 45
 50
 48
 60
 61


2
 8
 10
 15
 10
 19
 8
 13

a Data were generated using fractional occupations to assign case versus control status (IFRACT = 1) and excluded the training data when scoring and running 10 10-fold cross-validations for functions 1–5 (ITRAIN = 0, FRACT = 0.1, NTIME = 10). Training data is included in Additional file 1, Table S2.

b Simulated data sets were described previously [14] and were obtained from Dr. Moore by request.

c Total score is the summation over all scoring functions after linearly scaling the score for each individual function such that the top score is 50.0.

Mechanic et al. BMC Bioinformatics 2008 9:146   doi:10.1186/1471-2105-9-146