TITLE:
Prediction of Anti-Inflammatory Activity of a Series of Pyrimidine Derivatives, by Multiple Linear Regression and Artificial Neural Networks
AUTHORS:
Yafigui Traoré, Jean Missa Ehouman, Mamadou Guy-Richard Koné, Donourou Diabaté, Nahossé Ziao
KEYWORDS:
Anti-Inflammatory Activity, Multiple Linear Regression, Artificial Neural Network, QSAR
JOURNAL NAME:
Computational Chemistry,
Vol.10 No.4,
October
28,
2022
ABSTRACT: Anti-inflammatory activity of a series of tri-substituted pyrimidine derivatives was predicted using two Quantitative Structure-Activity Relationship models. These relationships were developed from molecular descriptors calculated using the DFT quantum chemistry method using the B3LYP/6-31G(d,p) level of theory and molecular lipophilicity. Thus, the four descriptors which are the dipole moment μD, the energy of the highest occupied molecular orbital EHOMO, the isotropic polarizability α and the ACD/logP lipophilicity were selected for this purpose. The Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) models are respectively accredited with the following statistical indicators: R2=91.28%, R2aj=89.11%, RMCE = 0.2831, R2ext=86.50% and R2=98.22%, R2aj=97.75%, RMCE = 0.1131, R2ext=98.54%. The results obtained with the artificial neural network are better than those of the multiple linear regression. However, these results show that the two models developed have very good predictive performance of anti-inflammatory activity. These two models can therefore be used to predict anti-inflammatory activity of new similar pyrimidine derivatives.