Advance Single Stage Convolutional Neural Network for Drug-Drug Interactions

Drug-drug interactions are one of the important and frequently happening medication issues in older age patients. The complexity of the drug-drug interactions and formation of side effects increases with a greater number of drugs consumed. Side effects are created when one drug's key factors or molecules inhibit or alter the functionality of the other drug. Drug - drug interactions can broadly be classified into pharmacokinetic and pharmacodynamics where both the states result in side effects, if not consumed with the right combination. Drug-Drug interactions happen because one drug interacts with other drugs during the consumption of both drugs at the same time. The current research talks about Convolutional neural networks with word embedding as input features. The goal of the papers is to propose a robust model that uses fewer parameters to train and arrive at a high accuracy model for predictions. We propose a novel single-stage detection method such as yolo or single shot detector to predict the output of the model. The object region will be treated as a simple traditional regression problem. The class probabilities of each region are calculated and predicted the bounding box coordinates. These models are used in predicting the class variables with high accuracy and much faster than a traditional two-stage detector models.