Significant research has been conducted in the area of developing in silico methods for predicting the affinity of drugs and drug-like compounds for human plasma proteins. The free fraction of a compound associated with a given level of binding affinity has a significant impact on the pharmacokinetic profile of a drug and its metabolites. The development of quality plasma protein binding models has become an important goal in assisting the drug optimization process. The structure, binding sites and binding interaction modes of common plasma proteins are discussed along with the protein composition of human plasma and pharmacokinetic consequences of protein binding profiles. A short section outlines current methods for measuring binding affinity for plasma proteins. A total of eighteen published studies were reviewed for this article and the statistical results from 42 models are tabulated and compared. Models are compared on the basis of the endpoint modeled, method of structure description, learning algorithm, validation criteria, and statistical results. The role of logP as an input descriptor and the possible utility of reported models in categorizing virtual compounds are also discussed.