The discovery and optimization of new drug candidates is becoming increasingly reliant upon the combination of experimental and computational approaches related to drug metabolism, toxicology and general biopharmaceutical properties. With the considerable output of high-throughput assays for cytochrome-P450-mediated drug–drug interactions, metabolic stability and assays for toxicology, we have orders of magnitude more data that will facilitate model building. A recursive partitioning model for human liver microsomal metabolic stability based on over 800 structurally diverse molecules was used to predict molecules with known log in vitro clearance data (Spearman's rho −0.64, P<0.0001). In addition, with solely published data, a quantitative structure–activity relationship for 66 inhibitors of the potassium channel human ether-a-gogo (hERG) that has been implicated in the failure of a number of recent drugs has been generated. This model has been validated with further published data for 25 molecules (Spearman's rho 0.83, P<0.0001). If continued value is to be realized from these types of computational models, there needs to be some applied research on their validation and optimization with new data. Some relatively simple approaches may have value when it comes to combining data from multiple models in order to improve and focus drug discovery on the molecules most likely to succeed.

This content is only available as a PDF.
You do not currently have access to this content.