Skip to main content
. 2015 Oct 13;9:26. doi: 10.1186/s40246-015-0048-9

Table 5.

Accuracy of CIMP prediction using gene expression profiles

Accuracy
Random Lasso Combined lasso Group lasso
Bladder 62.8 62.9 (p = 1) 74.2 (p≤2.10−16) 72.1 (p≤2.10−16)
Breast 80.5 83.9 (p≤2.10−16) 84.7 (p≤2.10−16) 85.5 (p≤2.10−16)
Colon 79.4 79.5 (p = 1) 95.0 (p≤2.10−16) 94.2 (p≤2.10−16)
Lung 73.2 84.2 (p≤2.10−16) 76.2 (p≤2.10−16) 86.6 (p≤2.10−16)
Stomach 57.6 81.2 (p≤2.10−16) 83.0 (p≤2.10−16) 84.8 (p≤2.10−16)
Overall 71.9 82.4 82.6 85.0

This table shows the accuracy, assessed by threefold cross-validation repeated 100 times over each tissue (first column), of sample classification in CIMP-positive and CIMP-negative classes from gene expression data using random classification (second column), lasso logistic regression (third column), combined lasso (fourth column), or group lasso logistic regression (fifth column)