Precise computational methods to determine ligand-binding affinity are needed to accelerate the discovery of new drugs. Assessing protein-ligand interaction is of great importance for virtual screening initiatives. The affinity may be computational evaluated using scoring functions involving terms for intermolecular hydrogen bonds, contact surface, hydrophobic contacts, electrostatic interactions and others. Empirical scoring functions have been developed to evaluate ligand-binding affinity very rapidly. In addition to predict affinity, these scoring functions have been employed to identify the best results obtained from docking simulations. This review describes several computational methods, employed to estimate ligand-binding affinity and discuss their development and main applications.