g

ecause the DOG algorithm has targeted both the up-regulated and

-regulated DEGs when outliers are present.

e 6.20 shows two heatmaps generated by the negative base ten

m of the p values of seven test models constructed for the same

d data set. In the heatmap shown in Figure 6.20(a), the t test,

ORT, DOG and OS models demonstrated a good performance

LSOSS and the MOST models, especially the LSOSS model,

how a good performance. As aforementioned, the COPA, OS and

orithms were all developed to deal with outliers presented in the

ated DEGs. Figure 6.20(b) shows a heatmap of the seven test

onstructed for the down-regulated DEGs. In this situation, only

model worked well and the COPA, OS and ORT models even

worse performance than the t test.

(a) (b)

The heatmaps of the negative base ten logarithm of the p values of seven tests

he toy data. (a) The heatmap for the up-regulated DEGs. (b) The heatmap for

egulated DEGs.

e 6.21(a) shows one up-regulated DEG. Because the outliers are

n the case replicates, the t test did not work very well, and its p

s 0.022. However, the COPA, OS, ORT and DOG models all

classified it as a DEG. In Figure 6.21(b), the gene expression