Toxicology Predictive Analytics
U.S. Environmental Protection Agency
Chemical testing is expensive and time consuming, and only a small fraction of chemicals have been fully evaluated for adverse effects on humans.
The U.S. Environmental Protection Agency (EPA) created a database of 1,800 chemicals that warranted detailed analysis. They wanted to accelerate testing of these chemicals by developing a predictive analytics solution that determines the lowest dose at which a change is observed in animal toxicity tests.
With an Analytics Starter Pack from Topcoder, the EPA received on-demand access to dozens of data scientists who worked to develop algorithmic solutions to the problem. Competitors used the EPA's database of chemicals as well as publicly available sources to predict minimum toxicity, and a public leaderboard helped drive competition.
The winning solution leveraged a novel approach that the EPA had not previously considered, and EPA scientists see a high potential for it to drive further advancement in the space. The results and findings of the project were presented at the 2014 EPA Toxcast Data Summit.