Table 1

Strengths and weaknesses of the hypergeometric test and individual components of RIDDLE

Method

Strengths

Weaknesses


Hypergeometric

Gold standard for overlapping sets

Fails for split sets


Local extension

Moderate improvement upon split sets


Diffusion - AUC

Exceptional improvement upon split sets

Considerable performance loss for GO overlap sets

Strong correlation with centrality


Diffusion - AP

Matches hypergeometric for overlapping sets

Major improvement upon split sets

Weak correlation with centrality

Strong correlation with set size


RIDDLE

Best for split sets

Minor performance loss for GO overlap sets


AP, average precision; AUC, area under the ROC curve; GO, Gene Ontology.

Wang et al. Genome Biology 2012 13:R125   doi:10.1186/gb-2012-13-12-r125

Open Data