Toward Dose\nPrediction at Point of Design

Human dose prediction (HDP) is a useful tool for compound\noptimization\nin preclinical drug discovery. We describe here our exclusively in\nsilico HDP strategy to triage compound designs for synthesis and experimental\nprofiling. Our goal is a model that provides a preliminary estimate\nof the dose for a given exposure target based on chemical structure.\nFirst, we construct machine learning models to estimate rat pharmacokinetics,\nwhich are subsequently allometrically scaled to estimate human pharmacokinetics.\nSecond, we establish a 10 nM free concentration target for early HDP\nwhere potency data are not yet available. Finally, we assess the uncertainty\nassociated with each model and propagate these into the final estimate,\nproviding us with actionable guidance on the level of accuracy of\nthese estimates. We find that this strategy can reduce preparation\nof compounds with poor properties relative to an unstructured approach,\nbut extensive experimental testing remains required.