Drug Discovery With Neural Networks

Discovering a new drug has always been a long process that takes years. With the recent advances of AI and the accumulation of research data in biological databases, the drug discovery process and the research pace is getting faster than ever. Researchers in the laboratory of innovation and science at Harvard are working on the Connectivity MAP project [1] with the goal of advancing drug development through improvements to the drugs MoA prediction algorithms. This challenge was launched as a kaggle competition [2] in order to build machine learning models to predict the MoA of unknown drugs. We start by understanding the competition's dataset: We have a dataset with gene expression and cell viability data as features and 206 MoA as targets.