DrugME

Inspiration

There were 1.5 million peer-reviewed papers published in 2010 and an estimated 2.5 - 5.6% growth in the total literature per year. Already, the amount of literature that a doctor has to review to keep up with the highest standard of care is strenuous and will soon approach impossibility. Our application addresses this exponential demand for clinicians to keep up with novel pharmacotherapies by automating and analyzing the vast amounts of relevant literature, then using it to ensure the most scientifically sound and personalized prescription in the clinic.

What it does

Clinical Decision Support Systems (CDSSs) are currently used by clinicians to assist in diagnostic and drug prescription. These solutions are not publicly available and, perhaps surprisingly, the databases are still updated by hand. This leads to inconsistencies introduced by human error and prescribes generalized and untailored therapeutic solutions to the patient. Our system employs a natural language parser that mines the literature to keep track of the gold standard therapies for specific diseases and takes into consideration personal patient information such as genetics, medical history, socioeconomic status to produce the most appropriate personalized drug therapies.

How we built it

We used OpenFDA to retrieve product monograph data for all 81000 drugs available. We then fed the textual data through NLTK, and along with some custom Natural Language Processing logic, determined the indications that the drug was responsible for treating. Finally, the CDSS interface was designed to fit into the workflows of physicians around the world.

Challenges we ran into

One of the main challenges we were faced with was the immense amount of drug data that was available from billions of sources. The amalgamation of this data proved to be a significant obstacle to our success since there was often repeated, and inconsistent data between the sources. Addressing these critical issues became essential to allowing for the success we have today.

Accomplishments that we're proud of

Processing and integrating 81000 drugs into a meaningful framework within a 36-hour timeframe.

What we learned

We did a significant amount of ideation on what the first iteration of our product will look like in the coming future. We have a much better view of the roadmap going forward.

What's next for DrugME

This project is to be continued to be developed at the University of Toronto under the Pharmacology and Biomedical Toxicology department. The parsing of cutting edge research articles to back up clinical data should be the next step in achieving our vision.

Built With

• express.js
• firebase
• ndc
• nltk
• openfda
• react
• redux