Biomedical Engineering Reference
In-Depth Information
projects have focused on less than 100 disease indications, yet the
standard disease ontologies have thousands to consider (for example,
Medical Subject Headings (MeSH) [6] has over 2850 diseases). To
consider all of the possibilities would be unrealistic without informatics.
In New Opportunities, the problem was simplifi ed and initially less than
150 additional indications were considered, which had clear unmet patient
needs. A system was therefore developed that leverages both the unstructured
data within SOLR (>50 M scientifi c documents) and the structured content
within an aggregated collection of competitive intelligence databases. This
brings together an understanding of biological and competitive rationale by
searching for sentence level co-occurrence between disease term(s) of
interest and the drug's mechanism term(s) as outlined in Figure 14.5.
This information is then stacked on top of each other in an interactive
heatmap visualisation. Figure 14.6 illustrates this approach and combines
this mining result with additional internal ideas. By providing a single
visualisation of over 30 000 serious opportunities, our scientists can
prioritise and evaluate new indications for our compounds, or identify
disease area themes to consider. Figure 14.7 shows an example of the
sentence level extraction that scientists are able to view when they drill
down into the results. In this way, compounds were identifi ed that had
potential across a range of eye and skin disorders. In 2009, a deal was
signed with Alcon, experts in ophthalmology, and in 2011 with Galderma,
experts in dermatology.
This system was built entirely using Pipeline Pilot [7], although there
are alternative workfl ow tools such as KNIME (see Chapter 6 by Meinl
and colleagues) and Taverna that could be considered. In this situation,
Pipeline Pilot provides extremely good fl exible web services that can be
Drug repurposing matrix. Schematic representing the
approach to search scientifi c literature and combine
with competitive intelligence for over 60 000 potential
combinations identifying hot spots to investigate further
Figure 14.5
Search WWH ::
Custom Search