Integrating Ligand and Target-Driven Based Virtual Screening Approaches With in vitro Human Cell Line Models and Time-Resolved Fluorescence Resonance Energy Transfer Assay to Identify Novel Hit Compounds Against BCL-2

• Published in: Frontiers in chemistry 2020-04, Vol.8, p.167-167
• Main Authors: Tutumlu, Gurbet, Dogan, Berna, Avsar, Timucin, Orhan, Muge Didem, Calis, Seyma, Durdagi, Serdar
• Format: Article
• Language: English
• Subjects: BCL-2; Chemistry; in vitro human cell line models; MD simulations; molecular docking; text mining; virtual screening
• ISSN: 2296-2646; 2296-2646
• DOI: 10.3389/fchem.2020.00167

Antiapoptotic members of - / -2 (BCL-2) family proteins are one of the overexpressed proteins in cancer cells that are oncogenic targets. As such, targeting of BCL-2 family proteins raises hopes for new therapeutic discoveries. Thus, we used multistep screening and filtering approaches that combine structure and ligand-based drug design to identify new, effective BCL-2 inhibitors from a small molecule database (Specs SC), which includes more than 210,000 compounds. This database is first filtered based on binary " model constructed with 886 training and 167 test set compounds and common 26 toxicity quantitative structure-activity relationships (QSAR) models. Predicted non-toxic compounds are considered for target-driven studies. Here, we applied two different approaches to filter and select hit compounds for further biological assays and human cell line experiments. In the first approach, a molecular docking and filtering approach is used to rank compounds based on their docking scores and only a few top-ranked molecules are selected for further long (100-ns) molecular dynamics (MD) simulations and tests. While docking algorithms are promising in predicting binding poses, they can be less prone to precisely predict ranking of compounds leading to decrease in the success rate of studies. Hence, in the second approach, top-docking poses of each compound filtered through QSAR studies are subjected to initially short (1 ns) MD simulations and their binding energies are calculated via molecular mechanics generalized Born surface area (MM/GBSA) method. Then, the compounds are ranked based on their average MM/GBSA energy values to select hit molecules for further long MD simulations and studies. Additionally, we have applied text-mining approaches to identify molecules that contain " " phrase as many of the approved drugs contain indole and indol derivatives. Around 2700 compounds are filtered based on " model and are then docked into BCL-2. Short MD simulations are performed for the top-docking poses for each compound in complex with BCL-2. The complexes are again ranked based on their MM/GBSA values to select hit molecules for further long MD simulations and studies. In total, seven molecules are subjected to biological activity tests in various human cancer cell lines as well as Time-Resolved Fluorescence Resonance Energy Transfer (TR-FRET) assay. Inhibitory concentrations are evaluated, and biological activities and apoptotic potentials are assessed by cell c