Over the last few decades, computer-aided drug design has emerged as a most successful technique rendering the drug discovery process more efficient and less costly. In the structure-based drug design branch, three-dimensional information on the biomolecular targets is used by the docking and scoring methodologies to find and optimize new ligands. In this dissertation, the following approaches are presented: ligand design, binding mode prediction, structure-activity relationship (SAR), molecular docking, receptor-ligand scoring and bridging water thermodynamics, followed by their application in protein-ligand complexes and host-guest systems. Molecular docking, which has become a powerful and influential tool for studying molecular recognition, aims to predict the binding mode of small molecules toward their biological target. It has been used in several projects: we have predicted the binding modes of a novel and potent inhibitor of CDK2 and FLT3 kinases and covalent inhibitors of AChE, BChE. Further, semi-empirical quantum mechanics-based scoring functions (SQM/COSMO) were used in native pose recognition where the poses had been generated with several docking programs. The SQM/COSMO scoring was compared with classical scoring function. We observed that SQM/COSMO accurately predicted the native poses in two dozen of difficult and diverse protein-ligand systems. Lastly, we discuss the important role of explicit water molecules in protein kinases, hydrolases, serine racemase and host-guest systems. We determined their thermodynamical parameter G which correlated very well with the experimental binding G for protein-ligand complex and host-guest systems.