Evaluation of co-administration in the treatment of patients with SARS-CoV-2 and drugs to treat comorbidities using artificial neural networks 388

Abstract

The aim of this work is to develop a clinical decision support tool to evaluate the result of the interaction of the drugs used to treat SARS-CoV-2 and the co-administration of drugs to treat comorbidities.

Neural networks (RNs) are used to predict drug interaction outcomes. The model has been developed using an artificial neural network of the PNN (Probabilistic Neural Network) / GRNN (General Regression Neural Network) type. Once the artificial neural network was trained, tested and validated, a determination coefficient of 82.79% was obtained, with which it is possible to have good and fast predictions to know the result of the interaction of specific drugs for SARS-CoV-2 with other drugs to treat different diseases.

For the training of the RN, a database was used, considering nine drugs to treat SARS-CoV-2 and 10 categories of drugs to treat other types of conditions. In the literature there are studies that demonstrate the importance of correct drug co-administration. However, there are currently no tools that integrate the use of neural networks and the new knowledge obtained from responses to drug interactions, which can be used first-hand by treatment specialists, decreasing with the risk of said interaction. (However, few efforts have been made to integrate artificial intelligence into this process.) Therefore, this paper aims to address this gap in knowledge. This system can be used as a decision support tool to evaluate and select between the best treatment to use in a patient with SARS-CoV-2, and avoid complications due to the negative interaction that it can cause with other medications.