GNN model for classification of SARS-CoV-2 severity in molecules

Abstract

In a time when healthcare issues and diseases get more complex every day, it becomes evident that efficient and precise disease detection and classification are invaluable. Through the quick development of machine learning methods and artificial intel ligence, there are now new and revolutionary techniques for disease detection and diagnosis. Conventional methods of detecting a disease may oversimplify this com plex relationship of dependence and reflection inside a bundle of dataset comprising extremely heterogeneous symptoms and pathologies. Therefore, conventional meth ods may fail to provide enough feedback and inputs to the medical unit. The main topic of this thesis is the usage of Graph Neural Networks (GNNs) to spot and di agnose diseases. Particularly, this analysis focuses on the ability of GNNs to assess COVID-19 severity based on the SMILES dataset. Particularly, this analysis focuses on the ability of GNNs to assess COVID-19 severity based on the SMILES dataset. This study proves that by exploiting the capacity of GNNs, GNNs can deliver the precision required for prompt interventions, and this results in improved patients’ outcomes and an effective healthcare system. The experimental results are highly promising, with GNNs achieving an accuracy of 87.16%, an F1 score of 82.63%, a precision of 84.27%, and a recall of 81.06% for Version 1 (not considering inactive cases), and an accuracy of 69.52%, an F1 score of 71.28%, a precision of 65.42%, and a recall of 78.30% for Version 2 (considering all cases — active, intermediate, and inactive). These data show that the GNNs approach is a successful method of classifying the level of severity of COVID-19 correctly by the way they depict the complicated connections of the dataset. This marks an ideal balance between the two metrics of precision and recall, suggesting that the model can correctly identify the cases and also minimize false negatives. This becomes even more important in a healthcare setting where the cost of misdiagnosis is extremely high. The article in general illustrates the capabilities of GNNs in transforming the process of disease diagnosis into a more efficient, effective, and accurate one, which can have a pro found meaning for doctors, patients, and other healthcare providers.