Predicting human subcutaneous bioavailability of monoclonal antibodies using an integrated in-vitro/in-silico approach

Monoclonal antibodies (mAbs) have become a cornerstone in therapeutic development, increasingly administered via subcutaneous (SC) injection due to its convenience and patient adherence benefits. However, accurately predicting SC bioavailability in humans remains a challenge, largely due to the limitations of traditional animal models that fail to provide reliable predictions for clinical outcomes, creating a significant gap in preclinical evaluations. To address this, we have developed an integrated in-vitro/in-silico approach that employs functional principal component analysis (FPCA) to summarize the release and transmission profiles information generated by the Subcutaneous Injection Site Simulator (SCISSOR) platform. The FPCA method extracted main shape functions from SCISSOR profiles, representing the most significant variations, and the resulting FPC scores were used as predictors in the modelling process. We employed self-validated ensemble modelling (SVEM) to predict the SC human bioavailability of mAbs based on the transmission and release features. SVEM is an ensemble modelling technique allowing the use of all observations for both training and validation making it a suitable method for small sample sizes. The model was further tested on new four commercial mAbs, demonstrating a good agreement between the predicted and actual bioavailability, and outperforming monkey data. We then elucidated how SCISSOR release and transmission profile are correlated with different mAbs and formulation parameters. This approach represents valuable addition to the toolkit for predicting the SC human bioavailability of mAbs. By combining in-vitro and in-silico methods, we offer a reliable approach that can outperform preclinical animal models.