The benefit offered by virtual screening methods during the early drug discovery process is directly related to the predictivity of scoring functions that assess protein-ligand binding affinity. The scoring of protein-ligand complexes, however, is still a challenge: despite great efforts, a universal and accurate scoring method has not been developed up to now. Targeted scoring functions, in contrast, enhance virtual screening performance significantly. This review analyzes recent developments and future directions in the area of targeted scoring functions.