“Pharmaceutical Knowledge Retrieval through Reasoning of chEMBL RDF” is
the title of my master thesis, a twenty-week research project performed at the
Department of Pharmaceutical
Bioscience at Uppsala University (Prof.
Wikberg, supervised by Egon
Willighagen). The project aims at using the ChEMBL data with a technology
that might be new to some: by using semantic web technologies. The
life sciences workbench Bioclipse
(doi:10.1186/1471-2105-10-397)
has support for several semantic web tools, including RDF, and was used to establish such a
connection.
Two aspects were looked at in this study. Firstly, we developed the search functionality for ChEMBL data to use RDF. For this, we took advantage of the RDF-ized ChEMBL knowledgebase (using the data from ChEMBL 02). Secondly, we developed a use case where compounds derived from ChEMBL are analyzed with the substructure mining software MoSS (see the Bioclipse Wiki). Here, we search for common and discriminative substructures within or between kinase families.
Within the context of these two aspects, we developed an application using both
the JavaScript and the
Wizard functionality in Bioclipse. The above shown wizard shows how various
searches for compound-protein interaction can be formulated. Results are shown
in the "Results table". The user can then select which data he wants to save, by
moving it to the lower table which lists the data that will be saved by this
wizard.
A second, more application-targeted Wizard was developed that primarily concentrates on retrieving compounds that bind proteins in a certain kinase family with a given activity type (see below). A histogram can be opened to visualize the distribution of activities. Lower and upper bound values can be selected, for focus, for example, only on that active compounds. A second, identical wizard page is provided to select a second dataset. This allows the user to set up a between-family data set. The saved data can then be used in the MoSS application to find the common and discriminative substructures (not shown).
Benefits of this approach focus on the data interoperability: the RDF
technologies are used as uniform and Open Standard access to the ChEMBL data.
Using this approach, implementing new search queries is very easy, and does not
require one to know anything about the database schema; a common controlled
vocabulary (ontology) hides those implementation details. Community standards
for such vocabularies are under development, and will integrating the ChEMBL
data with other databases and other applications.
Does this sounds interesting to you, or do like to give us feedback? Please send a note to annzi.andersson+chembl@gmail.com . Further details are provided in my blog!
Sincerely, Annsofie Andersson.
Two aspects were looked at in this study. Firstly, we developed the search functionality for ChEMBL data to use RDF. For this, we took advantage of the RDF-ized ChEMBL knowledgebase (using the data from ChEMBL 02). Secondly, we developed a use case where compounds derived from ChEMBL are analyzed with the substructure mining software MoSS (see the Bioclipse Wiki). Here, we search for common and discriminative substructures within or between kinase families.
A second, more application-targeted Wizard was developed that primarily concentrates on retrieving compounds that bind proteins in a certain kinase family with a given activity type (see below). A histogram can be opened to visualize the distribution of activities. Lower and upper bound values can be selected, for focus, for example, only on that active compounds. A second, identical wizard page is provided to select a second dataset. This allows the user to set up a between-family data set. The saved data can then be used in the MoSS application to find the common and discriminative substructures (not shown).
Does this sounds interesting to you, or do like to give us feedback? Please send a note to annzi.andersson+chembl@gmail.com . Further details are provided in my blog!
Sincerely, Annsofie Andersson.
Comments