Biomedical Engineering Reference
In-Depth Information
non-homogeneous probe eects (i.e., dierent expression intensities among
a set of probes). Such non-homogeneous probe eects shown in probe-level
expressions may not be reected in gene-level summary data (see Fig-
ure 1). Moreover, expression dierences in a two-group comparison could
vary among a set of probes. Some probes may have large dierential expres-
sions whereas some probes yield similar expressions between the two groups.
The dependency of dierential expressions on probes indicates an interac-
tion eect between probe and treatment eects. The interaction eect may
have potential biological implications, such as alternative splicing
35
. In this
case, gene-level data analysis may miss this target gene. Even if the gene is
identied, without the information of probe expressions, it is hard to judge
the occurrence of alternative splicing. In this chapter, we present a probe
rank approach to analyze probe level data.
Distribution of Probe Expression
(Probe Set ID= 1007_s_at in a HU133 2.0 Plus Gene Chip)
RMA
Median
Mean
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
probe id
Fig. 1. Heterogeneity of probe eects in a given gene expression. The probe intensity
shows large variation, ranging from 100 to 1100 with median 392 (denoted by the dashed
line) and geometric mean 376 (denoted by the dotted line). The gene expression using
RMA is 321 (denoted by the solid line).
Search WWH ::
Custom Search