NewDrugTargets.org is an interactive visualization tool for
discovering interesting associations between diseases and
potential drug targets.
We used natural language processing to identify disease and
protein mentions in the text of
PubMed
abstracts. Using this
data, we derived two metrics: novelty and
importance. Novelty measures the relative scarcity of
specific publications about a given concept (such as a target
or a disease), while importance measures the relative
strength of the association between two concepts. We then
built this web tool, which enables users to explore the
relationships between the novelty of potential drug targets
and their importance to diseases.
Our approach was guided by the following assumptions:
A target that is mentioned in many abstracts that also
mention a specific disease is likely to be of
importance to that disease.
A target or disease that is mentioned in fewer
abstracts is more novel and less well understood.
Abstracts which mention only a few targets and diseases
are more specific and should be given greater weight
than those which mention many.
For more details about how these scores are computed, please
see this poster.