Machine learning-based QSAR modeling, molecular docking, dynamics simulation studies for cytotoxicity prediction in MDA-MB231 triple-negative breast cancer cell line

Abstract

Triple-negative breast cancer (TNBC) is the most aggressive cancer type that tests negative for progesterone, estrogen, and HER2 protein receptors. Therefore, TNBC is unlikely to respond either to drugs that target HER2 receptors or to hormonal therapy drugs. In the present study, 2D-QSAR model was developed to predict the inhibitory concentration (IC50) of terpene derivatives against human breast cancer metastatic cell line MDA-MB231, using V-Life MDS v4.5, module. The model was developed using the forward stepwise multiple linear regression method, with a regression coefficient (r2) of 0.86, and a cross-validated r2 (q2) of 0.8448. Molecular descriptors namely electronegativity (Epsilon-3), carbon atoms separated through five bond distances (T_C_C_5), Sum of Electrotopological state indices of –CH2 group (SssCH2count), Dipole moment of coordinate-z (Zcomp Dipole), and Distance between highest positive and negative electrostatic potential on van-der Waals surface area (Most +ve & -ve Potential Distance) were found to be the most potent contributing chemical descriptors for cytotoxicity against TNBC MDA-MB231 cell line. Additionally, binding affinities and interaction patterns revealed that the proposed compounds exhibit good binding affinities and substantial stability with c-Met-and β-tubulin receptors, as assessed by docking studies. Molecular dynamics simulations (100 ns) and binding free energy calculations were also performed using the MMGBSA method. The pharmacokinetic and eADME/T analysis of predicted compounds were assessed through Discovery Studio software. These findings may be of immense importance in the optimization and development of dual inhibitory potent anti-cancer inhibitors against the MDA-MB231 TNBC cell line. Moreover, this novel QSAR based prediction model/method is implemented in R-package and software tool developed for virtual screening purposes and available online for download through the public repository GitHub.