The dissertation presents the application and development of a method for G protein-coupled receptors’ homology modeling, and the use of the models in virtual screening. The adopted method consists in optimization of the binding site, using the structure of an active ligand. Studies were carried out on the dopamine D1 receptor, which is an example of a target with marked influence on biological systems and therefore of high pharmacological importance. The effectiveness of the method for constructing receptor models and providing accurate prediction of ligand-receptor interactions was confirmed by assessing the impact of its key elements on the results of retrospective virtual screening trials. The novel models have shown superior predictive ability in virtual screening, compared with the reference models, constructed based on previously-used methodology. The novel approach has also proved to be more effective than the method based on the properties of molecules. Thus, it has been demonstrated that the novel models correctly describe the affinity of compounds (particularly agonists), structurally distant from the ligands used in the process of model optimization. In addition, the present work has established that the use of a set of chemically diverse ligands at the stage of model optimization, as well as an effective method for selecting their best representatives, is crucial for quality of the models. A thorough analysis of the D1 receptor binding site and binding mode of the ligands was performed, in the context of selectivity versus D2 receptor binding. Finally, a prospective virtual screening of an in-house chemical compound library was carried out and enabled identification of a new chemotype of D1 receptor ligands. High affinity of a representative compound from this library was confirmed by in vitro studies.