Skip to main content

Advertisement

Table 4 Performance statistics of selected genes on MS data

From: Weighted-SAMGSR: combining significance analysis of microarray-gene set reduction algorithm with pathway topology-based weights to select relevant genes

  Training set (10-fold CV results) Test set
A. Performance comparison
 Method (n) Error (%) GBS BCM AUPR Error (%) GBS BCM AUPR
 SAMGSR (52) 34.09 0.244 0.570 0.645 46.67 0.465 0.501 0.725
 W-SAMGSR (25) 31.82 0.191 0.611 0.771 43.33 0.341 0.564 0.860
 LASSO (30) 34.09 0.275 0.632 0.672 46.67 0.377 0.499 0.747
 Penalized SVM(11) 47.73 0.406 0.534 0.630 45 0.569 0.431 0.555
 gelnet (169) 34.09 0.251 0.528 0.589 46.67 0.246 0.547 0.746
 RRFE (198) 43.18 0.263 0.547 0.619 46.67 0.300 0.523 0.693
B. Performance of the top 3 teams in sbv MS sub-challenge (among 54 teams)
 Study (size) Training data used/Method used Error (%) GBS BCM AUPR
 Lauria’s (n > 100) E-MTAB-69/Mann-Whitney test, then use top α % of the selected genes and Cytoscape to get the clusters on the test set -- -- 0.884 0.874
 Tarca’s (n = 2) GSE21942 (on Human Gene 1.0 ST)/LDA -- -- 0.629 0.819
 Zhao’s (n = 58) 7 other data and E-MTAB-69/Elastic net 30 -- 0.576 0.820
  1. Note: W-SAMGSR weighted-SAMGSR, LDA linear discrimination analysis, gelnet generalized elastic net by [25], RRFE reweighted recursive feature elimination by [14]
  2. --: not available. Lauria’s Tarca’s and Zhao’s studies [38, 39, 44] are the 3 best studies in the sbv MS sub-challenge