Pharma insight on digital marketing, social media, mobile apps, online video, websites and interactive healthcare tools
by Dominic Tyer
Kaggle challenge set up to help pharma company develop safe and effective medicines
Merck & Co has become the latest pharma company to bring gamification to its clinical research by tapping into online scientific research community Kaggle.
Merck launched the Merck Molecular Activity Challenge, with its $40,000 prize fund, in August to assist its drug development work by predicting molecular activity.
The competition was won by a team of three US academics, who beat 15 other teams to identify the best statistical technique for predicting biological activity of different molecules.
It was based on 15 molecular activity data sets, each for a biologically relevant target and each row corresponded to a molecule and contains descriptors derived from that molecule's chemical structure.
The first place prize of $22,000 went to a team of academics from the University of Toronto and the University of Washington who have expertise in defining the state-of-the-art in machine learning.
The Kaggle community numbers around 31,000 data scientists who work across disciplines and industries to tackle problems outside their realm of expertise.
It runs predictive modeling competitions for companies, governments and researchers who set problems and supply the prize money in exchange for the intellectual property behind the winning model
Kaggle has previously been used by NASA and Ford to solve problems, and earlier this year Boehringer Ingelheim become the first pharmaceutical company to use it, launching a biological response competition.
Merck's molecular data competition winners
The winning team in Merck's competition was made up of doctoral students George Dahl, Navdeep Jaitly and Christopher Jordan-Squire, and assistant professor in statistics and computer science Ruslan Salakhutdinov.
They share the first place prize of $22,000 after triumphing by using the competition to illustrate the ability of neural network models to perform well with no feature engineering and only minimal pre-processing
The second place team of Xavier Conort, Jeremy Achin and Tom DeGodoy included two of the scientists that won the Boehringer competition (Achin and DeGodoy) and shares $10,000.
Merck also ran a visualisation challenge with a $2,000 prize for the most insightful and elegant graphical representations of the data, and this was won by Laurens van der Maaten, a post-doctoral research at Delft University of Technology in The Netherlands.