Reliability of Virtual Screening Methods in Prediction of PDE4B-inhibitor Activity

Identification of active ligands using computational methods is a challenging task. For example, molecular docking, pharmacophore modeling, and three dimensional quantitative structure-activity relationship models (3D-QSAR) are widely used methods to identify novel small molecules. However, all these methods have, in addition to advantages, also significant pitfalls. The aim of this study was to compare some commonly used computational methods to estimate their ability to separate highly active PDE4B-inhibitors from less active and inactive ones. Here, 152 molecules with pIC50-range of 3.4-10.5, originating from six original studies were used. High correlation coefficients by using docking, docking with postprocessing with molecular mechanics-generalized Born-surface area -method (MMGBSA), pharmacophore modeling, and 3D-QSAR were obtained. These results are well in line with earlier studies done with similar methods, and suggest that computational methods could be successfully used to identify novel PDE4B-inhibitors, especially if using multiple methods together.