Model-Based Virtual PK/PD Exploration and Machine Learning Approach to Define PK Drivers in Early Drug Discovery.

While poor translatability of preclinical efficacy models can be responsible for clinical phase II failures, misdefinition of the optimal PK properties required to achieve therapeutic efficacy can also be a contributing factor. In the present work, the pharmacological dependency of PK end points in driving efficacy is demonstrated for six common pharmacological processes via model-based analysis. The analysis shows that the response is driven by multiple pharmacology-specific PK end points that change with how the response is defined. Moreover, the results demonstrate that the most important chemical structural features influencing response are specific to both target and downstream pharmacology, meaning the design and screening criteria must be defined uniquely for each target and corresponding pharmacology. The model-based virtual exploration of PK/PD relationships presented in this work offers one approach to identify target pharmacology-specific PK drivers and the associated potency-ADME space early in discovery to increase the probability of success and, ultimately, clinical attrition.