Title: Docking study and QSAR analysis based on the artificial neural network and multiple linear regression of novel harmine derivatives

Authors: Taoufik Akabli; Hamid Toufik; Mourad Stitou; Fatima Lamchouri

Addresses: Laboratory of Natural Substances, Pharmacology, Environment, Modeling, Health and Quality of Life (SNAMOPEQ), Polydisciplinary Faculty of Taza, Sidi Mohamed Ben Abdellah University of Fez, Morocco ' Laboratory of Natural Substances, Pharmacology, Environment, Modeling, Health and Quality of Life (SNAMOPEQ), Polydisciplinary Faculty of Taza, Sidi Mohamed Ben Abdellah University of Fez, Morocco ' Laboratory of Natural Substances, Pharmacology, Environment, Modeling, Health and Quality of Life (SNAMOPEQ), Polydisciplinary Faculty of Taza, Sidi Mohamed Ben Abdellah University of Fez, Morocco ' Laboratory of Natural Substances, Pharmacology, Environment, Modeling, Health and Quality of Life (SNAMOPEQ), Polydisciplinary Faculty of Taza, Sidi Mohamed Ben Abdellah University of Fez, Morocco

Abstract: Harmine and its derivatives are an important class of natural molecules for fighting cancer. Researching for the physical characteristics involved in this activity provides crucial keys to develop new derivatives which are more active and less toxic. For this purpose, a series of 50 harmine derivatives were studied using molecular modelling, namely 2D-QSAR analysis and molecular docking. The best 2D-QSAR model was developed correlating the three most important descriptors with the cytotoxic activity using MLR and ANN. The statistical analysis indicates high performance of the established models (R²MLR = 0.77, q²MLR = 0.73, R²_extMLR = 0.81, Q²F3MLR = 0.70, r²m_MLR = 0.71 and CCC_MLR = 0.88, R²ANN = 0.86, q²ANN = 0.79 and R²extANN = 0.76). The analysis of the selected three descriptors showed that the lipophilicity remains the crucial property on which cytotoxic activity depends. Moreover, molecular docking of the most active compound (44) shows that it takes up a good pose into the active site of DYRKA1 kinase, as reflected by the low binding energy (-10.5 kcal/mol) and the various interactions formed with the amino acids. Thus, these results were exploited to design six new derivatives having high predicted pIC50, low binding energy and exciting pharmacokinetics properties.

Keywords: ADME/Tox properties; artificial neural network; ANN; descriptors; lipophilicity; multiple linear regression; MLR; molecular modelling.

DOI: 10.1504/IJCAET.2023.127797

International Journal of Computer Aided Engineering and Technology, 2023 Vol.18 No.1/2/3, pp.190 - 210

Received: 30 Mar 2020
Accepted: 26 Jun 2020
Published online: 19 Dec 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article