Efficient Prediction Adverse Drug-Drug Interactions with Deep Neural Networks

These days, combining many medications is the best course of treatment to slow the pathologic process, which includes a number of underlying negative effects brought on by drug-drug interactions (DDIs). The evaluation of pharmacological interactions, pharmacodynamics, and probable adverse effects using artificial intelligence (AI) is a possibility. Many AI-based DDI prediction methods, including both machine learning and deep learning, that make use of the available huge data have been described in recent years. Even if past techniques yielded notable advancements, changes are still essential. In this study, we present the results of three distinct single models: Random Forest (RF), Recurrent Neural Networks (RNNs), and Deep Neural Network (DNN) in our initial filtration round to find a suitable candidate model to predict the risk or severity of negative consequences can be elevated when drug A is combined with drug B. After evaluating the models' performance. This paper suggested DNN as a way to enhance DDIs' predictive capabilities. The prediction algorithm was 96.2% accurate in predicting 86 different kinds of DDIs using a benchmark dataset. The proposed model applies pre-processing techniques on data because these techniques align with our proposed model goals of data-driven decision-making and actionable insights in prediction of DDI. After applying the processing to the dataset, the suggested model DNN classifier outperforms the already proposed approaches on the same dataset. Integrating sustainable principles into DDI prediction using DNNs entails creating models that not only achieve high accuracy but also operate efficiently, reducing the environmental impact of large-scale data processing. The great performance of our model positions it at the top of the list of those well-designed pharmacovigilance-assisted tools that make it easier to find DDIs to support clinical judgment and drug development.