A machine learning method for predicting the IC50 values of novel designed analogs of Non-Nucleoside Reverse-Transcriptase Inhibitors (NNRTIs) as potentially safer drugs

Abstract

It is estimated that over four decades since the onset of the AIDS epidemic, a significant number of individuals have been impacted by HIV-1 infection. The latest data shows that 165,000 children were among the 1.5 million people who contracted HIV in 2022. Additionally, it is projected that about 655,000 people will die from AIDS-related illnesses in 2023. Scientists have been working on different methodologies to improve the pharmacological properties of drugs, one of which is rational drug design. To build a robust machine-learning model, this study used computer-aided drug design and virtual screening strategies based on ligand and structure. This allowed us to select the best scaffold for modifying the core of nevirapine. Based on this screening, forty analogs were built, and the prediction of IC50, affinity, and toxicity was performed. The study found that changes should be made to the activated pyridine ring, specifically in the alpha position to the pyridine nitrogen.Moreover, the substitution with the best in silico values was the meta substitution in the carbocyclic ring. It was suggested that these substituents can be either electro-donating or electron-withdrawing groups. This QSAR-SBDD analysis provides a roadmap for synthesizing novel NNRTIs with improved pharmacological profiles.