Towards A Computational QSP Platform To Support Antibody Drug Conjugate Design In Cancer

Objectives: Antibody-drug conjugates (ADCs) are designed to deliver cytotoxic drugs to specific cancer cells but their development is hindered by complex interactions between antibody, linker, and payload. These interactions determine an ADC's distribution and efficacy. To accelerate development by predicting outcomes and optimizing trial design, we propose to leverage the data from approved and under investigation ADCs with the development of a QSP platform for ADCs, able to handle heterogeneous data and non-linear interactions. Here, we showcase initial developments towards such a platform using PADCEV, a Nectin-4 targeting ADC with Monomethyl auristatin E (MMAE) as payload, as an example. Methods: We developed a multi-species physiologically-based pharmacokinetic (PBPK) model to simulate ADC and free payload distribution. In order to reproduce the training pharmacokinetic (PK) data set (mouse data for free MMAE and human data for PADCEV), we implemented payload release through either on target release or non specific clearance. We connected this to a model of ADC binding and internalization. To circumvent the non-availability of key in vitro data on PADCEV, we made use of data on brentuximab vedotin (BV), another ADC with the same linker and payload, to inform this part of the model. Then, we coupled this model to a Simeoni tumor growth model which reproduces preclinical tumor volume profiles of bladder cancer [1]. Finally, we trained this resulting ADC model with PADCEV efficacy data (disease progression), from mice experiments and from a Phase 1 study [2]. For the latter, we generated a large realistic virtual population and sampled virtual patients to reproduce Phase 1 outcomes.Results: The platform, informed by PADCEV data supplemented with in vitro BV data, captures well PADCEV’s PK (free payload, total antibody and ADC) and efficacy with the reproduction of a preclinical experiment and a Phase I study. As an illustrative example of the platform’s capabilities, we simulated a scenario where drug developers are faced with two linker choices with different stability. For this, we compared the two options in terms of both PK and efficacy and found that with low stability linkers, ADC exposure was reduced while increasing the free payload concentration. This reduces on-target efficacy while increasing off-target effect. Conclusion: Here, we report the initial developments of a QSP platform for ADC. We show in this proof of concept how the ADC platform applied to PADCEV captures the reported PK and efficacy and how it can be used to answer questions typically arising during ADC development such as optimal linker stability. This platform has the potential to significantly reduce the time and cost of ADC development, and future applications could include the integration of more ADCs, adverse effects or more cancer types.Citations: [1] P. M. Challita-Eid et al. Enfortumab Vedotin Antibody–Drug Conjugate Targeting Nectin-4 Is a Highly Potent Therapeutic Agent in Multiple Preclinical Cancer Models. Cancer Res. (2016).[2] J. Rosenberg et al. EV-101: A Phase I Study of Single-Agent Enfortumab Vedotin in Patients With Nectin-4–Positive Solid Tumors, Including Metastatic Urothelial Carcinoma. J Clin Oncol (2020).