_

g

_ ,

ur models. For instance, the differential expressions were 0.019

for 1552816_at, 2.91 and 0.632 for 1556328_at. Generally

when the number of replicates is smaller, it is certain that

variance information is lost causing difficulty in DEG discovery

e 6.10. Genes identified as DEGs by DSG only or agreed by four models.

DEGs detected by DSG only

GSM242136

GSM242196

GSM242207

GSM242208

at

– 0.5713881

3.180547

2.753406

1.924831

t

3.0230166

1.599343

– 1.007124

2.519831

DEGs detected by all four models

GSM242136

GSM242196

GSM242207

GSM242208

at

3.078238

3.800072

3.0971330

1.9006288

at

2.701267

1.110215

– 0.2126983

0.4778602

y

pter has discussed several issues of the gene expression pattern

y. Importantly, this chapter has discussed the issues which

y occurred in biological/medical gene expression data analysis.

the outlier problem, the bimodality problem and the insufficient

problem. For each of these issues, several methods or algorithms

n introduced and compared in this chapter. However, none can be

superior compared with the others so far. There is still a great

improving these methods or developing new algorithms in the

or instance, the DSG is required to be extended to multi-

nal space gene expression modelling based on the current work.

required to have a unified approach for detecting heterogenous

d bimodal genes.