A comparative evaluation of data-merging and meta-analysis methods for reconstructing gene-gene interactions

BMC Bioinformatics

Table 3 Reconstruction of Ikaros regulatory program in PBMC data collection. For each method the number of predicted and correctly retrieved interactions is reported, along with the odds ratio, precision and recall performances (see text for further details on these metrics)

Method	# Predicted interactions	# Retrieved interactions	Odds ratio	Odds ratio significance	Precision	Recall
Combat	82	21	2.2726	0.00022	0.2561	0.0093
Fixed-Effect	102	21	1.827	0.00440	0.2059	0.0093
Single-Dataset	387	70	1.6513	[<0.0001 - 0.65914]	0.1861	0.0310
No-Correction	113	21	1.6491	0.01435	0.185841	0.0092

Odds ratio statistical significance is assessed through the hypergeometric test. For the Single-Dataset approach distinct performances and significance p-values were computed for each dataset, summarized here as a weighted average of the performance and with the interval spanned by the p-values, respectively

ISSN: 1471-2105