The distribution of the numbers of the bimodal genes identified by the bimodal
n ten runs for the letrozole drug data.
mulated data analysis
was composed of 1,000 simulated genes with 40 replicates.
hese simulated genes, 100 were designed as bimodal genes. In
20 genes (ten unimodal and ten bimodal) were composed of
m two outliers. Table 6.9 shows four confusion matrices for the
ihood ratio) test, the GM (gap maximisation) test and the BI
index) test. Figure 6.41 shows the ROC curves for three test
GM, LR and BI). The GM and the BI tests worked well compared
LR test.
The confusion matrices of the LR, BI and GM tests run on the simulated data
the gene expression bimodality pattern discovery simulation. UN stands for
BM stands for bimodal. MCC stands for the Mathews correlation coefficient
1977].
GM
LR
BI
UN
BM
UN
BM
UN
BM
UN
898
2
900
0
900
0
BM
0
100
94
6
10
90
MCC
0.9836
0.2120
0.9377
e 6.42 shows the logarithm gene expression profile of a designed
gene. The p value of the GM test was 0.0065. However, the LR
d to identify it. The bimodal index test also failed to identify this
gene. Its bimodal index test value was 0.69, but the threshold
separating the bimodal genes from the unimodal genes for this
mulated data set was 1.84.
