MLR و ANN دراسة العلاقة الكمية بين البنية و الكيمياء الكهربية لإزاحة نصف الموجة التأكسدية لمركبات الإندوليزين بإستخدام الحركية الدوائية

Abstract

This study aims to establish a quantitative relationship between the structure of 52 indolizine derivatives and their electrochemical property represented by the oxidation half-wave potential (E1/2), using the QSER method. Indolizine is a heterocyclic compound with a fused structure combining pyrrole and pyridine rings, which imparts various biological activities such as antimicrobial, anti-inflammatory, antioxidant, and anticancer effects. The research employed a range of computational methodologies to study these compounds, using several techniques to calculate molecular descriptors with the assistance of software tools including HyperChem, GaussView, and the SwissADME web server. A predictive mathematical model was constructed using Multiple Linear Regression (MLR), which produced acceptable results based on the following statistical parameters: R² = 79.46 ; 𝑄𝐿𝑂𝑂 = 69.24 and 𝑄𝑒𝑥𝑡 62.37. The model was then further developed using Artificial Neural Networks (ANN), yielding improved statistical values: R² = 85.48 ; 𝑄𝐿𝑂𝑂 =69.95 and 𝑄𝑒𝑥𝑡 94.83. A comparison indicates that the ANN outperforms the MLR in terms of statistical accuracy. To assess the pharmacokinetic potential of the studied compounds, an analysis of their ADME properties was conducted using the SwissADME web server.