Advancing QSAR models in drug discovery for best practices, theoretical foundations, and applications in targeting nuclear factor-κB inhibitors- A bright future in pharmaceutical chemistry

Abstract

Developing robust and valuable quantitative structure-activity relationship (QSAR) models has become increasingly significant in modern drug design. These models play a crucial role by enabling the determination of molecular properties of compounds and predicting their bioactivities for therapeutic targets. QSAR models utilize various machine learning methods, such as support vector machines (SVM), multiple linear regression (MLR), and artificial neural networks (ANNs). These widely applicable methods have substantial implications for developing more precise medicines. The effectiveness of QSAR research dramatically relies on how each process step is conducted and how the analysis is carried out. This paper discusses the essential steps in developing and validating QSAR models using machine learning. A case study is presented to provide a clear example, focusing on 121 compounds acting as potent nuclear factor-κB inhibitors (NF-κB). The study compares multiple predictive QSAR models based primarily on linear and non-linear regression techniques.