Detecting Side Effect of Drug Molecules Using Recurrent Neural Networks

N. Nalini, Shaik.irham saleha, Katam Jagan Mohan Reddy, Panchagiri Raju, K Sasi Kiran

doi:10.64751/

Authors

N. Nalini, Shaik.irham saleha, Katam Jagan Mohan Reddy, Panchagiri Raju, K Sasi Kiran Author

DOI:

https://doi.org/10.64751/

Keywords:

Adverse drug reaction, drug-drug interaction, side effect prediction, graph neural network, self-supervised learning, scientific machine learning

Abstract

ADRs, bad reactions to drugs, which occur when drugs react with one another, are a significant health issue in the world population as they lead to mortality, morbidity and expensive medical care. The problem is aggravated by the fact that therapeutics becomes more complex, and the population is ageing as well. Currently, ADRs cannot be discovered in a normal manner until patients report it when the drugs are already on market. The Two Sides Drug Bank collection is applied in this case, which contains much information on how drugs interlink with one another and what their side effects are. KNN and DT are examples of the traditional methods that have been utilized to find ADRs. Nevertheless, these models are not effective at identifying complicated trends in the data. To circumvent this issue, more precise methods and GNN are applied to locate ADRs by considering drug interactions as graphs. The accuracy of 99.74 when feature extraction is done by 2D CNN is a significant performance enhancement. This is a better approach than the previous algorithms, and it appears to be a promising approach to identifying ADRs at an early stage and improving the situation with the state of health.