A Data Warehousing Approach for Genomics Data Meta-Analysis - Evolving Application Domains of Data Warehousing and Mining - page 152

In summary , in this section, we have first
described which synthetic data are stored and se-
mantically annotated in the AMI data warehouse.
We have presented the underlying ontologies that
support the annotation process.

knowledge of biologists. These annotations are
detailed in the subsections below.

Annotation of Scientific Papers

Here is an example of a sentence extracted from
a scientific paper:

SeMAntIc WeB technologIeS
for MetA-AnAlySeS “In vitro assays demonstrated that only p38alpha

and p38beta are inhibited by csaids.” Our purpose
is to extract, from this sentence, relevant bio-
medical concepts and the relationship expressed
between these concepts.

One of the originalities of the AMI approach is
to take advantage of semantic annotations, about
expression data and related information, for que-
rying the AMI knowledge base. This section is
organized in two subsections. The first subsection
details the different kinds of annotations in AMI
and the second subsection shows how they are
involved in a semantic search among data.

AMI reuses here an existing module called
MeatAnnot (Khelif et al., 2007). Starting from a
scientific paper, MeatAnnot generates a structured
annotation, based on the MeatOnto ontology,
which describes interactions between genes/pro-
teins and other concepts. The annotation process
consists on three steps:

Semantic Annotations

Annotating a resource tends to associate the most
relevant descriptive information to it. Here we de-
tail the AMI approach for annotating the different
types of resources (see AMI Annotation&Storing
Tools in Figure 1). This includes annotations on
scientific papers, on micro-array experiments, on
statistic and data mining results and on the implicit

(i) Instances of MeatOnto concepts and re-

lationships are detected in the sentence:
'' p38alpha '' and '' p38beta '' are detected
as instances of the concept '' Gene_or_
Genome '', '' csaids '' is detected as instance of
the concept '' Pharmacologic_Substance '',

Figure 10. Fragment of GMineOnto