Structure-based drug design (SBDD) is widely used in lead optimization to predict binding affinity of virtual compounds proposed for synthesis. However, it does not address other factors that determine <i>in vivo</i> efficacy such as bioavailability or pharmacokinetics. Previously, we developed a methodology to predict the bioavailability (%F) of molecules within a compound series using physiologically-based pharmacokinetics (PBPK) simulations. This approach requires a small number of rat bioavailability measurements taken during lead optimization to train a local QSAR model for an apparent intrinsic clearance. In addition, a quadratic partial least squares regression (PLSR) model is fitted to thousands of PBPK simulations of related drug-like compounds to characterize the “bioavailability landscape” of the series. The PLSR model coefficients describe global sensitivity analysis (GSA) to identify sensitive compound properties to the change in %F.<br/><br/>The PLSR models approximate the full PBPK rat %F well for the series and the models are computationally efficient. These PLSR models have been integrated into Molsoft ICM molecular modeling software to enable multi-parameter optimization (MPO) functions that combine binding affinity and bioavailability. These can be used with manual interactive SBDD guided by the sensitive properties from the GSA. MPO can also be applied to combinatorial virtual libraries to identify promising compounds for synthesis. Due to the combinatorial growth of virtual libraries, efficient optimization and evolutionary algorithms have been applied to search for compounds with both good affinity and bioavailability.

Rational bioavailability optimization using physiologically-based pharmacokinetics simulations

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.