Group-based quantitative structure-activity relationship (G-QSAR) and molecular docking studies of BCL-XL, BCL-2 and PI3K-I inhibitors for lung cancer treatment

Abstract

Lung cancer is one of the common cancers in Malaysia and there are many different types of drugs used in the treatment of lung cancer. Bcl-xL, Bcl-2 and PI3K-γ were found to play important role in lung cancer cell proliferation and development. Bcl-xL and Bcl-2 allow uncontrolled cell proliferation by inhibiting apoptosis while derangement of PI3K-γ pathway has been associated with uncontrolled cellular proliferation and survival. Thus these proteins become good targets for the development of targeted therapy in lung cancer. Various classes of compounds have been used successfully to targets Bcl-xL, Bcl-2 and PI3K-γ and in order to improve the biological activity of these compounds various QSAR methods have been employed. In this study, an advanced method in computational drug design, Group-based Quantity Structural Activity Relationship (G-QSAR) was performed using V-LIFE® on dataset of non-congeneric compounds with binding activity available in Binding Database website from various literatures and generated potential protein inhibitors. Combination of variable selection methods (simulated annealing, stepwise forward or stepwise forward backward) with model building method (multiple linear regression, K-nearest neighbour method or partial least square) has come out with three statistical significant models for each protein target which were used to generate new inhibitor candidates by computational method. G-QSAR analysis of Bcl-xL inhibitors has resulted in model SA-MLR (r²=0.78, q²=0.68), STP-MLR (r²=0.80, q²=0.70) and SA-kNN (q²=0.75) which have been used to generate total number of 2520, 3600 and 5850 compounds respectively. Three best models obtained from Bcl-2 inhibitors G-QSAR analysis are STP-MLR (r²=0.82, q²=0.71), SA-MLR (r²=0.83, q²=0.75) and SA-kNN (q²=0.65), which generated 4964, 1536 and 7576 compounds respectively. While G-QSAR analysis of PI3K-γ inhibitors come out with STP-MLR (r²=0.72, q²=0.67), SA-MLR (r²=0.70, q²=0.65) and STP-PLS (r²=0.68, q²=0.62), which generated 2040, 2244 and 2560 compounds respectively. Newly generated compounds were then analysed and validated by molecular docking into the available Bcl-xL, Bcl-2 and PI3K-γ crystal structures from Protein Databank using Glide® software. Six series of docking (docking of three datasets of G-QSAR models-based generated compounds and three docking of known target protein inhibitors) were performed for each protein target. High throughput virtual screening (HTVS) docking was conducted to filter large compound dataset, and this was followed by extra precision (XP) docking which is more detail and flexible than HTVS. Overall, XP docking results of Bcl-xL inhibitors showed common interaction with ASN136, ARG139, GLY138 and TYR101. Most of the Bcl-2 inhibitors interacted with Bcl-2 crystal structure at amino acid residues LEU96, ALA108, PHE112, PHE71 and TYR67. Lastly, majority of the PI3K-γ docked inhibitors were shown to interact with four common amino acids including TYR867, LYS833, VAL882 and TRP812. Non-covalent interactions such as hydrophobic, π-π interaction and hydrogen bond are the common types of interaction found in docking results of Bcl-xL, Bcl-2 and PI3K-γ inhibitors. Based on the docking results, the newly generated compounds interacted very well with target proteins as good as the known inhibitors in clinical trials.