A Data Warehousing Approach for Genomics Data Meta-Analysis - Evolving Application Domains of Data Warehousing and Mining - page 158

Figure 16. Example of a CORESE query using XPATH embedded in SPARQL

XPATH Embedded in
SPARQL: An Example

• The function xpath evaluates the XPATH

expression “//PMML/.../Cluster[contai
ns(Genes,'CCL19')...]...” : it returns the
list of all clusters that contain both genes
'CCL19' and 'CCL5'. The search process
is performed on the PMML documents
( ?pmmlURI }) referred in the clusters
annotations.
The SPARQL query filter on clusters

Descriptions of gene clusters are stored in a PMML
document referred by semantic annotations. So it
is easier to answer to the query “Which molecular
functions are associated to clusters containing
both genes CCL19 and CCL5 “ by combining
the information stored into annotations and into
PMML documents. This can be done via XPATH
code embedded in SPARQL through the semantic
search engine CORESE.
It was introduced in the semantic search
engine CORESE, a new function xpath with
two arguments. The first argument is the URI of
an XML document. The second argument is an
XPath expression. The purpose of this function
is to evaluate the XPath expression on the given
XML document.
Figure 16 shows the CORESE query that
matches the question: “which molecular functions
are associated to clusters containing both genes
CCL19 and CCL5” .
In this example:

•

( ?cluster ) contained on the list returned by
the XPATH query (Filter(xsd:string(?clust
er)=xpath(...)).

The result is a list of triples. Each triple con-
tains:

• the cluster URI ( ?cluster ) (e.g.

Cluster_GSE6281_hclust_1 ),
the experiment identifier ( • ?expe ) attached

to the cluster (e.g. GSE6281 ),
the biological functions ( • ?molfunc ) associ-

ated to the cluster (e.g. cytoskeleton ).

In summary , in this section, after a first
section that details semantic annotations, we
have shown how SQL and XPATH functions
associated with the CORESE search engine give
to AMI the ability to provide semantic answers
and inferred knowledge while incorporating all
related original data.

• The SPARQL query retrieves all exist-

ing clusters ( ?cluster ) and the molecular
functions ( ?biofunc ), experiment identifier
( ?expe ) and PMML URI( ?pmmlURI) asso-
ciated to each retrieved cluster.