Skip to main content

Table 3 Evaluation of the overlap among different approaches to prioritize classifier's mechanisms

From: Curation-free biomodules mechanisms in prostate cancer predict recurrent disease

Approaches to Prioritize Classifier's Mechanisms Overlapping Mechanism (#) % of each compared set Overlapping p-value (FET)
FAIME-Derived Approaches Overlap-Gene Ontology    
  Yu ∩ Wallace 462 15%; 38% 6.9 x10-6
  Yu ∩ Taylor 824 26%;34% 3.8 x10-3
  Wallace ∩ Taylor 360 29%;15% 2.2 x10-5
  Yu ∩ Wallace ∩ Taylor 137 4%;11%;6%  
  Overlap Cancer Module    
  Yu ∩ Wallace 41 15%;49% 1.9x10-4
  Yu ∩ Taylor 77 27%;46% 5.7 x10-6
  Wallace ∩ Taylor 25 30%;15% 4.8 x10-4
  Yu ∩ Wallace ∩ Taylor 14 5%;17%;8%  
Controls: Standard Bioinformatics Approaches Overlap of Differentially Expressed Genes    
  Yu ∩ Wallace 33 11%;6% 6.1 x10-9
  Yu ∩ Taylor 65 22%;24% <2 x10-16
  Wallace ∩ Taylor 20 3%;8% 1.5 x10-3
  Yu ∩ Wallace ∩ Taylor 7 2%;1%;3%  
  Overlap GO derived from Differentially Expressed Genes    
  Yu ∩ Wallace 1 5%;10% 4.2 x10-2
  Yu ∩ Taylor 0 NA NS
  Wallace ∩ Taylor 0 NA NS
  Yu ∩ Wallace ∩ Taylor 0 NA NS
  1. Legend: NA = not applicable; NS = not significant; #= count of significant mechanisms overlap