Molecular activity prediction using deep learning software library
Yoshiki Kato, Shinji Hamada, Midori Goto · 2016
In order to know how work deep learning method in chemoinformatics and bioinformatics problems, we have attempted to predict the molecular activities using the molecular fingerprints (chemical descriptor vectors) provided by the “Merck molecular activity challenge” competition and an open source deep learning library Chainer. Our result has been able to reproduce almost identical increase-decrease tendencies with the correlation Rs2of the champion group in the competition. GPU performance was also examined and the speed gain were more than 11 times than only CPU computation.