Retrospective Validation of a Structure-Based Virtual Screening Protocol to Identify Ligands for Estrogen Receptor Alpha and Its Application to Identify the Alpha-Mangostin Binding Pose

Abstract

The publicly available enhanced data of ligands and decoys for estrogen receptor alpha (ERα) which were recently published has made the retrospective validation of a structure-based virtual screening (SBVS) protocol to

Binary Quantitative Structure-Activity Relationship Analysis in Retrospective Structure-Based Virtual Screening Campaigns Targeting Estrogen Receptor Alpha

“…Although the predictive ability of SBVS_1 has already outperformed the original SBVS campaigns accompanying DUD-E [23] and our previous retrospective SBVS campaigns using PLANTS1.2 [14], the predictive abilities of SBVS_2 and SBVS_3 are considerably better than SBVS_1.…”

“…compound as the objective function to rank the compounds and select compounds within 1% false positives (FP) as positive (SBVS_1) [14], (ii) using decision tree resulted from the binary QSAR analysis using ChemPLP score and PLIF bitstrings of the best pose of every compound as descriptors (SBVS_2) [26], and (iii) using decision tree resulted from the binary QSAR analysis using ensPLIF of the selected poses from optimized ChemPLP score as the cutoff (SBVS_3). Although the predictive ability of SBVS_1 has already outperformed the original SBVS campaigns accompanying DUD-E [23] and our previous retrospective SBVS campaigns using PLANTS1.2 [14], the predictive abilities of SBVS_2 and SBVS_3 are considerably better than SBVS_1.…”

“…The distance between the PLIF of the predicted pose and the PLIF of the reference pose calculated using Tanimoto metric results in Tc-IFP [1] or Tc-PLIF [7], which could be used as alternative scoring functions in structure-based virtual screening (SBVS) campaigns [3,4,[12][13][14][15]. Notably, this scoring function is a reference-dependent function, and the selection of the reference determines the predictive quality of the SBVS protocol [3,16].…”

“…Previous attempts with the mol2 formats from DUD-E [23] showed that employing decision tree resulted from binary quantitative structure-activity relationship (QSAR) analysis using recursive partition and regression tree method (RPART) [24], and ChemPLP score as the docking score and PLIF bitstring as the descriptors had significantly better predictive ability to identify potent ERα ligands compared to the protocol that using only ChemPLP score [14]. Notably, instead of using the readily to be docked three-dimensional (3D) formats of compounds provided by DUD-E [23], in the research presented in this article, SMILES format was selected to avoid bias in ligand preparation steps [22,25].…”

Binary Quantitative Structure-Activity Relationship Analysis to Increase the Predictive Ability of Structure-Based Virtual Screening Campaigns Targeting Cyclooxygenase-2

“…The standard retrospective SBVS campaigns using PLANTS1.2 followed by PyPLIF [18,41] targeting COX-2 [25] using ligands and decoys from DUD-E [27] resulted in 3,536,850 docking poses. There were 900 docking poses missing because the docking software could not simulate 6 out of 23,150 decoys.…”