A multiscale quantitative systems pharmacology platform for early screening of MMAE-based antibody-drug conjugates

Antibody-drug conjugates (ADCs) have become an important therapeutic modality in the cancer field. However, their complex structural characteristics bring huge challenges to ADC development. Quantitative systems pharmacology (QSP) models can help integrate multi-scale data, providing mechanistic support. Nevertheless, the application of this tool in the early drug discovery stage remains limited. We developed and validated a multi-scale QSP model that integrates systemic pharmacokinetics (PK), single-cell disposition and tumor distribution models, quantitatively predicting intratumoral unconjugated payload exposure and spatial distribution. This model was applied to three valine-citrulline-monomethyl auristatin E (vc-MMAE) based ADCs targeting different antigens (RC48, RC108, and RC118). The data from xenograft mice were used for model validation. The results showed that the prediction error range for intratumoral free monomethyl auristatin E (MMAE) exposure was between -38.13% and 20.52%, indicating acceptable predictive accuracy. Based on this validated QSP model, the distribution of unconjugated payload in tumors was simulated for the three ADCs at a dose of 1.5 mg/kg. By combining the predicted exposure and distribution of unconjugated payload in tumors with in vitro cytotoxicity (IC₅₀), we successfully reproduced the observed in vivo efficacy ranking of these three ADCs. Global sensitivity analysis indicated that intratumoral payload exposure is mainly driven by target density, ADC permeability, and the physicochemical properties of the payload, such as cellular uptake rate, efflux rate, and payload permeability. This model allows for the replacement of key parameters with human-related data, which is not achievable in animal experiments. Ultimately, this work established a mechanism-based in silico tool for early ADC candidate screening and structural optimization.