A Data Warehousing Approach for Genomics Data Meta-Analysis - Evolving Application Domains of Data Warehousing and Mining - page 131

Table 1. Gene expression data from a micro-array

Probe Symbol c 1 c 2 … … c j … … … c n

probe 1 gene 1 exp 11 exp 12 … … exp 1j … … … exp 1n

probe 2 gene 2 exp 21 exp 22 … … exp 2j … … … exp 2n

… … … … … … … … … … …

probe i … … exp i2 … … exp ij … … … exp in

… … … … … … … … … … …

probe p … … exp p2 … … exp pj … … … exp pn

gene expression data

The chapter is organized in seven sections:
the second section “MICRO-ARRAYS EX-
PERIMENTS” presents micro-array experiment
basic principles and guidelines of micro-array
data processing and analysis. The third section
“AMI OVERVIEW” details our motivations in
designing the AMI approach and the framework
of this semantic data warehouse. In the fourth
section “STANDARD META-ANALYSES OF
DNA MICRO-ARRAYS”, we discuss standard
meta-analyses approaches and we provide some
experimental results that justify the AMI data
warehouse oriented approach. The fifth section
“EXPRESSION DATA, SYNTHETIC DATA
AND METADATA” details data storage and
representation in AMI; it presents XML-based
representations of statistical results on micro-array
gene expression data and metadata describing
them. In the sixth section “SEMANTIC WEB
TECHNOLOGIES FOR META-ANALYSES”,
we show how semantic web technologies provide
a useful insight into the wide diversity of the data
warehouse. We conclude in the seventh section.

Micro-arrays are arrays of genetic material. This
technology consists on printing DNA brands (DNA
sequences) on a chip of plastic or glass and making
them hybridize with similar radioactively labeled
DNA sequences ( probes ) extracted from a tissue
of interest. The number of hybridizations gives
an estimation of the gene expression for the gene
matching the probe hybridizedAffymetrix 9 micro-
arrays that become a standard, use a technique that
synthesizes probes in situ while in two channels
micro-arrays technology, DNA sequences are
directly deposed on the chip. Whatever the kind
of chip, three steps are followed: labeling with
fluorescent dyes, hybridization and image scan-
ning.After probes are hybridized, the light emitted
is captured by a scanner and the chip produces an
image that produces a result in its turn since the
image is processed to extract numeric data.
Gene expression data may be seen as the Table
1 where each column c j represents a particular
condition or sample (tissue, kinetic…). Generally
one column at least represents a control condi-
tion. Each item exp ij represents the expression of
a probe probe i with its matching gene i under the
condition c j . Some columns may refer to so called
replicates that measure gene expression in similar
conditions to ensure more robustness. Raw gene
expression data need to be pre-processed before
statistical and data mining analyses.

MIcro-ArrAyS exPerIMentS

In this section, we introduce main concepts and
standards of micro-array domain and we give
some insights in most frequent methods applied
to analyze their data.