Investigation of the Anticancer Potential of 2-alkoxycarbonylallyl Esters Against Metastatic Murine Breast Cancer Line 4T1 Targeting the EGFR: A Combined Molecular Docking, QSAR, and Machine Learning Approach

Abstract

Background: The search for novel and potent anticancer drugs is imperative. This present study aims to unravel the mechanisms of action of 2-alkoxyecarbonyl esters using robust model(s) that can accurately predict the bioactivity of novel compounds. Twenty-four potential anticancer 2- alkoxycarbonylallyl ester compounds obtained from the literature were employed in building a 3D-QSAR model. Objectives: The objective of this study is to determine the predictive ability of the GFA-based QSAR models and extreme machine learning models and compare them. The lead compounds and newly designed compounds were docked at the active site of a human epidermal growth factor receptor (EGFR) kinase domain to determine their binding modes and affinity. Methods: QikProp program and Spartan packages were employed for screening compounds for druglikeness and toxicity. QSAR models were equally used to predict the bioactivities of these molecules using the Material Studio package. Molecular docking of the molecules at the active site of an EGFR receptor, 1M17, was done using Auto dock tools. Results: The model of choice, with r2pred (0.857), satisfied the recommended standard for a stable and reliable model. The low value of r2, Q2 for several trials and cRp2 (0.779 ≥ 0.5) and the high value of correlation coefficient r2 for the training set (0.918) and test set (0.849) provide credence to the predictability of the model. The superior inhibition of EGFR displayed by the lead compounds (20 and 21) with binding energies of 6.70 and 7.00 kcalmol-1, respectively, is likely due to the presence of double bonds and α-ester groups. ADMET screening showed that these compounds are highly druggable. The designed compounds (A and B) displayed better inhibition of EGFR. Conclusion: The QSAR model used here performed better than the Random Forest Regression model for predicting the bioactivity of these anticancer compounds, while the designed compounds (A and B) performed better with higher binding affinity than the lead compounds. Implementing the developed model would be helpful in the search for novel anticancer agents.