Alternative Transcript Analysis Methods for Exon Arrays
Revision Date: 2005-10-11
Revision Version: 1.1
Affymetrix GeneChip® Exon Array Whitepaper Collection:
1 of 13
Alternative Transcript Analysis Methods for Exon Arrays
I. Introduction
With exon arrays like the GeneChip® Human Exon 1.0 ST Array, researchers
can examine the transcriptional profile of an entire gene. Being able to gather
data for each individual exon enables the investigation of phenomena such as
alternative splicing, alternative promoter usage, and alternative termination while
also providing more probe level data to determine the overall level of expression
a particular locus. Detecting relative changes in alternative transcript forms is the
subject of this whitepaper.
We conclude that an ANOVA based method applied to a Splicing Index equal to
the ratio of exon signal to gene signal works well in two test data sets: one a
tissue panel and the other a set of cancerous and normal samples from the same
tissue. For the tissue panel data set, a correlation based method, Robust PAC,
performs well.
One of the most well studied alternative splicing events is the alternative
utilization of “cassette exons”. Here, we describe the evaluation of the
performance of different mathematical methods designed to detect alternatively
spliced cassette exons from the exon array data.
To do this, we first tested the method’s robustness in an artificial data set by
constructing “genes” comprised of a group of probe selection regions that are
best suggested by experimentally-derived annotation evidence as biologically
real and are always jointly expressed in a sample data set. Each such group of
probe selection regions is referred to as the “core constitutive exons” of that
“gene.” We then simulated alternative splice events in each such “gene” by
substituting exons from alternative regions of the genome and constructed ROC
curves as a way to measure the performance of the different methods.
II. Data set
II.A. Core Constitutive Ex