(a) (b)

a) The ROC curves and the AUC values for the two-dimensional simulated data

y four algorithms. (b) The Venn diagram of the predicted DEGs from four

the two-dimensional simulated data.

eal data set study

et was used to analyse breast cancer cell lines to examine how

expression inhibits metastatic cell invasion from lung cancer

breast cells (GDS3138) [Tavazoie, et al., 2008]. For this data set,

es, the cyberT and the SAMr algorithms were used in addition to

algorithm for the DEG discovery. Figure 6.52 shows the Venn

of comparing four models for the data about the numbers of the

ed DEGs. The result shows a great deviation among four models.

bers of discovered DEGs were 4542, 11794, 879, and 925 by the

berT model, SAMr and eBayes models, respectively. Only 660

ere discovered by all four models.

e 6.53 shows the distribution of DEGs discovered by four models.

seen that the SAMr model and the eBayes model demonstrated a

attern of DEGs, the cyberT model had too many discovered DEGs

t extended towards two sides. The DSG model discovered DEGs

ating gene differential expression between two means, which is

the fold change. This is why it had a uniform boundary on both