Plasma protein binding of drugs has been studied for almost 100 years, but despite the accumulation of large amounts of data, the accurate prediction of this ADME parameter continues to be problematic. This review outlines recent efforts on the development of prediction tools for plasma protein binding of drugs, specifically human serum albumin, in the context of its relevance and its influencing factors. The issue of why it is difficult to achieve prediction of sufficient quality for a diverse dataset will also be considered.