Challenges for in silico modeling of ADME data

COMP 195

Terry R Stouch, Computer-Assisted Drug Design, Computer-Assisted Drug Design, Bristol-Myers Squibb, MS H23-07, PO Box 4000, Princeton, NJ 08543-4000
The approximations inherent in much ADME data presents several challenges for in silico modeling. The properties, themselves, are often not rigorously defined and are often multicomponent with difficult underlying biology. Endpoints are not always determined precisely and are often qualitative, making quantitation difficult. Demands on these models can be extreme and go beyond good QSAR practices commonly employed for less challenging, but still demanding, endpoints. In particular, "general" models that cover large amounts of chemical space are desirable. Yet, the lack of standardization of assays means that data can not be quantitatively compared between laboratories, and often can not be compared at all, making it difficult to assemble sufficient data to develop such models. These and other problems will be detailed and their effect on the resulting modeling effort will be discussed