Breaking Barriers Between Species: Integrated Multi-Species PK Models Improve Human PK Predictions

The accurate prediction of human pharmacokinetics (PK) is critical for compound selection and determining the human efficacious dose in drug discovery. Besides predicting the key PK parameters clearance (CL) and steady-state distribution volume (V_ss), the prediction of exposure metrics, e.g., maximum and trough concentrations (C_max, C_trough), are essential to evaluate efficacy and toxicity. Still, methods for predicting further PK-shape-defining parameters like absorption rate constant (k_a), central and peripheral distribution volume (V_c, V_p), and intercompartmental clearance (Q) remain underexplored. Thus, this work aimed to improve the prediction of respective parameters, building up on a published PK-prediction-framework. The updated framework leveraged a multi-species PK model, enabling the joint prediction of shape-defining parameters by incorporating allometric scaling, while keeping established mechanism-based prediction methods for bioavailability (F) and CL. The performance of the multi-species model was tested for 40 small molecules with known human PK profiles and benchmarked against the literature framework, published alongside the data. The new predictions showed a reduction in geometric mean fold-error (GMFE) compared to the benchmark for both the predicted PK profile and its shape, associated with an improved prediction of Q and late concentrations (C_24h). The overall improvement was primarily driven by compounds, exhibiting inconsistent PK model structures across species when applying the benchmark method. Meanwhile, the prediction performance on other compounds remained comparable, making the integrated multi-species PK-model framework a promising approach for human PK prediction, especially in cases where difficulties in individual species model integration are faced or accurate trough concentration prediction is a priority.