Molecular docking is a computational screening approach in drug design able to predict the conformation of a protein-ligand complex. Docking algorithms provide an efficient and cost-effective alternative to experimental high-throughput screenings. Molecular docking is generally applied starting from the knowledge of the protein binding region. However, a precise information about the correct binding site is often missing and it becomes necessary to explore the entire protein surface by docking algorithms. Several methods have been developed to overcome the problem of not recognizing the binding site. In the present thesis, a new methodology to identify the experimental binding mode of small molecule ligands into protein structures where the real binding sites are unknown will be presented.