Predicting plasma protein binding of drugs: a new approach.

In spite of the large amount of plasma protein binding data for drugs, it is not obvious and there is no clear consensus among different disciplines how to deal with this parameter in multidimensional lead optimization strategies. In this work, we have made a comprehensive study on the importance of plasma protein binding and the influencing factors in order to get new insights for this molecular property. Our analysis of the distribution of percentage plasma protein binding among therapeutic drugs showed that no general rules for protein binding can be derived, except for the class of chemotherapeutics, where a clear trend towards lower binding could be observed. For the majority of indication areas, however, empirical rules are missing. We present here an extensive list of multiply determined primary association constants for binding to human serum albumin (HSA) for 138 compounds from the literature. Correlating these binding constants with the percentage fraction of protein bound showed that the percentage data above 90%, corresponding to a binding constant below 6 microM, are of insufficient accuracy. Furthermore, it could be demonstrated that the lipophilicity of drugs, traditionally felt to dominate binding to HSA, is not the only relevant descriptor. Here, we report a generic model for the prediction of drug association constants to HSA, which uses a pharmacophoric similarity concept and partial least square analysis (PLS) to construct a quantitative structure-activity relationship. It is able to single out the submicromolar to nanomolar binders, i.e. to differentiate between 99.0 and 99.99% plasma protein binding. Depending on the system, this can be important in medicinal chemistry programs and may together with other computed physicochemical and ADME properties assist in the prioritization of synthetic strategies.