Search for: [Abstract = "erature data. The PDE4B model, developed on the basis of the available crystal structures, also showed the interaction mode of the selective ligand characterized by the highest inhibitory activity. Optimization of the protein model, docking reference inhibitors by the induced\-fit method, retrospective virtual screening, and molecular dynamics simulations were used to prepare and verify all structural models. The analysis of MD simulations contributed also to the generation of specific pharmacophore models, the so\-called dynophores. The developed pharmacophores and structural models were applied to perform a two\-stage process of virtual screening of 4,126,936 ligands\: matching to pharmacophore hypotheses and docking of ligand groups selected in the first stage. In the ligand\-based approach, the conducted studies consisted of the development of empirical regression models using machine learning, which aimed to predict the value of pIC50 inhibitory activity. These models were trained on molecular descriptors describing the structures of known ligands of biological targets. In the process of virtual screening, their use will allow the selection of potentially effective inhibitors and provide information about the presumed IC50 value. Candidates for multi\-target\-directed ligands selected in both approaches were analyzed in terms of assigned values of individual scoring functions a"]
