EMPLOYING RECURSIVE PARTITION AND REGRESSION TREE METHOD TO INCREASE THE QUALITY OF STRUCTURE-BASED VIRTUAL SCREENING IN THE ESTROGEN RECEPTOR ALPHA LIGANDS IDENTIFICATION
Enade Perdana Istyastono · Asian Journal of Pharmaceutical and Clinical Research · 2015
Objective: Increase the predictive quality of the structure-based virtual screening (SBVS) protocol to identify potent ligands for estrogen receptor alpha (ERα). Methods: Employing recursive partition and regression tree (RPART) method to identify potent ligands for ERα among their decoys by using molecular docking scores and the protein-ligand interaction fingerprint bitstrings as the predictors. These predictors were obtained from previously published SBVS campaign to identify potent ligands for ERα. The quality of the protocol by using RPART method was assessed by examining the enrichment factors and the accuracy in 95% level of confidence compared to the reference protocol. Results: The decision tree resulted from analysis using RPART method increased the enrichment factor and the accuracy values of the SBVS protocol from 18.5 to 247.9 and from 0.975 to 0.989, respectively. Notably, the accuracy value of the protocol using the decision tree was statistically significant in 95% level of confidence while the reference protocol was not. Conclusion: RPART method could lead to a significant increase of the SBVS quality to identify potent ligands for ERα. Keywords: Recursive partition and regression tree, Molecular docking, Interaction fingerprint, Estrogen receptor alpha.