Mechanism of Action: A Comparative Study

In this problem, we predict the mechanism of action (MoA) of different drugs, which refers to how a drug produces its effect at molecular, cellular or physiological level. The aim is to predict multiple targets of MoA responses for thousands of drugs, based on features such as gene expression, cell viability, control perturbation type, time and dose, respectively. The dataset that we have used consists of over 200 targets of enzymes and receptors and over 5000 drugs. This is a multi-label classification problem. We have used various algorithms like random forest, XGBoost and neural network. We have evaluated the performance of different algorithms based on their log loss metric to determine the optimal model for our dataset.