Estimation of bioavailability and toxicity at the very beginning of the drug development process is one of the big challenges in drug discovery. Most of the processes involved in ADME are driven by rather unspecific interactions between drugs and biological macromolecules. Within the past decade, drug transport pumps such as P-glycoprotein (Pgp) have gained increasing interest in the early ADME profiling process. Due to the high structural diversity of ligands of Pgp, traditional QSAR methods were only successful within analogous series of compounds. We used an approach based on similarity calculations to predict Pgp-inhibitory activity of a series of propafenone analogues. This SIBAR approach is based on selection of a highly diverse reference compound set and calculation of similarity values to these reference compounds. The similarity values (denoted as SIBAR descriptors) are then used for PLS analysis. Our results show, that for a set of 131 propafenone type compounds, models with good predictivity were obtained both in cross validation procedures and with a 31-compound external test set. Thus, these new descriptors might be a versatile tool for generation of predictive ADME models.