Information Technology Reference
In-Depth Information
Fig. 14.1
An example of alternative splicing. A gene consists of coding regions (
dark
exons) and
non-coding regions (
white
introns). Exons are combined together by splicing mechanism to create a
normal transcript. In case of alternative splicing, transcripts are created with different use of exons
create functional units, proteins. This process of collecting exons is called
splicing
in genomics.
Alternative splicing
refers to the events of using exons differently than normal
cases in splicing. Figure
14.1
shows a pictorial example of alternative splicing. Clas-
sical gene microarrays have probes to measure the amount of transcripts (mRNAs)
created as a result of splicing, and therefore cannot capture alternative splicing events.
Exon microarrays can capture alternative splicing. Different usage of exons can be
caused by several reasons including mutations in exon sequences, and it has impli-
cations in development of cancers, for example.
Exon microarrays measure the expression level of individual exons, and these are
grouped as genes as we can see in Fig.
14.1
. So it would make sense to select all
exons that belong to the same gene when they are relevant. But one may also want to
assign different weights on exons, to detect alternative splicing events if any exists.
Selecting individual exons without considering their grouping as genes also remains
as an option, but it is more likely to overfit given data when
p
is large since it is not
constrained by group information.
14.3.1 Data
A combination of two microarray data sets available at Gene Expression Omnibus
(GEO),
3
with accession numbers GSE21713 and GSE32664 [
9
,
19
], have been used
in our case study. These contain total 113 microarrays from neuroblastoma patients.
Neuroblastoma is one of the most common solid cancer in children who are usually
younger than two years. Both GEO data sets have been obtained using the same
microarray platform, the Affymetrix Human Exon v1.0 ST arrays.
As up to four probes are used to measure the expression of a single exon, raw
measurements in microarrays have to be summarized and normalized. For these
we apply the frozen RMA (fRMA) algorithm by [
14
], which processes individual
microarrays using information from predefined global reference arrays. Low quality
Search WWH ::
Custom Search