Two-way AIC: detection of differentially expressed genes from large scale microarray meta-dataset

BMC Genomics

Table 1 Results of comparisons of each method's performance

Method	$\bar{s e}$	$\bar{s p}$	$\bar{p}$	$\bar{n d}$	$\bar{p d}$
1. two-way AIC	0.58578	0.99998	2.721 × 10^-5	5.71280	0.10295
2. t-rank/F -test	0.58477	0.99821	7.901 × 10^-3	245	4.41521
3. RankProducts	0.58597	0.99717	1.123 × 10^-2	312	5.62263
4. SAM	0.58690	0.99983	9.034 × 10^-4	96	1.73004
5. U -value (gene side)	0.65665	0.68416	2.085 × 10^-1	54.49481	0.98206
6. U -value (experiment side)	0.75034	0.99967	5.325 × 10^-4	23.91349	0.43095
7. 2-σ	0.65270	0.99871	5.202 × 10^-3	74.96886	1.35103
8. 3-σ	0.65488	0.99990	4.030 × 10^-4	17.01730	0.30667

We assess the performance of all methods by calculating $\bar{s e}$ (average sensitivity of detection of gene pairs in all operons), $\bar{s p}$ (average specificity of detection of gene pairs in all operons), $\bar{p}$ (average p-values of detection of gene pairs in all operons), $\bar{n d}$ (number of DEGs), and $\bar{p d}$ (the percentage of DEGs). Threshold of t-rank/F-test, RankProducts and SAM is set to 245, 312 and 96 to match the sensitivity with that of two-way AIC.

ISSN: 1471-2164