The rate of new medicines being discovered is decreasing exponentially. At the same time, the the cost and time involved with bringing a new drug to market is _ increasing _ exponentially. We were inspired by the potential of A.I. and big data to develop an innovative solution to save time and money for developing new drugs to save lives.

What it does offers a simple web platform where scientists can upload chemical data, and then receive predictions for the likelihood of a potential molecule binding to it's biological target.

How we built it

Josh engineered the ML algorithm from scratch and hosted it through flask on Google App Engine, Jasmin and Aman did a fantastic job on the front end using bootstrap, raw html and css, plus a little jquery here and there, all hosted of course, on our Flask server, and Aakaash solved challenging middleware problems as well as working on the front end.

Challenges we ran into

Our data set of bio assays had a huge class imbalance, 1 active molecule to ~200 non-activate, which we had to deal with using cost sensitive learning.

Hosting our ML algorithm online was also a mission, we had to set up a virtual environment and go through trail and error of setting up Flask with our dependancies.

Accomplishments that we're proud of

Developing an accurate ML model that is hosted online, easily accessible through a beautiful web interface.

What we learned

How to take ML to the web!

What's next for

Refining algorithm, and implementing new features to flesh out into a more robust health care platform.

Share this project: