A large number of molecules are generated using modern drug discovery tools such as combinatorial chemistry. The high costs involved in ADME (Absorption, Distribution, Metabolism and Excretion) assays highlights the importance of the development of in silico models, which are readily available for a number of pharmacokinetic and physicochemical properties. In order to obtain such predictive models it is required to generate a data set including a large number of compounds associated to the corresponding experimental data. Plasma protein binding (PPB) is an important pharmacokinetic property in drug design. The drug is transported in blood stream bound to plasma protein and only the free drug can bind to the molecular target and exert the pharmacological and physiological effects. PPB also affects the bioavailability because the fraction of bound drug is not available for target modulation. Hologram QSAR (HQSAR) is a technique that employs specialized fragment fingerprints (molecular holograms) to predict the property of interest. In the present work, we describe the results of the development of a predictive pharmacokinetic model for PPB. The data set consisting of 342 molecules collected from the literature with PPB values ranging from 0 – 99.9% was divided into training set (277 molecules) and test set (65 molecules). The best statistical HQSAR model (q² = 0.70, r² = 0.82, with 8 optimum components) was obtained using the fragment distinction Atoms, Bonds, Hydrogen and Donor & Acceptor, with holograms length of 257 and fragment size 2-5. The external validation of the model was carried out employing a test set and the results were in good agreement with experimental values. Basically, 75% of molecules with standard deviation (SD) smaller than 10%, whereas 19% with SD between 11 and 20% and other 6% with SD above 20%. These results show that the generated model is useful for the prediction of PPB at early stages of drug design.