How in silico structural properties help predict aberrant Fc-fusion pharmacokinetics
A new joint study between Amgen and COSBI has been published in The AAPS Journal, presenting an innovative approach to predict pharmacokinetic behavior of Fc-fusion proteins from in silico structural properties.
The paper, titled “Predicting Aberrant Fc-fusion Protein Pharmacokinetics from In Silico Structural Properties and Physiologically Based Pharmacokinetic (PBPK) Modeling”, is authored by Danilo Tomasoni, Alessio Paris, Roberto Visintainer, Kevin D. Cook, Aochiu Chen, Isabel Figueroa, Veena A. Thomas, and Luca Marchetti.
The work addresses a critical challenge in the development of Fc-fusion biologics. While fusion to the Fc domain is typically used to improve pharmacokinetics, some candidates still exhibit unexpectedly rapid clearance, often associated with non-specific off-target binding.
To tackle this issue, the study introduces an AI-assisted framework that extends a classical physiologically based pharmacokinetic (PBPK) model. The model incorporates two scaling factors (F1 and F2), derived as analytical functions of in silico physicochemical properties of proteins.
These analytical relationships were identified using SymbolicR, a novel symbolic regression framework developed to extract explicit, interpretable formulas directly from data. The resulting model was evaluated on an independent validation set, demonstrating its ability to generalize to new Fc-fusion proteins within the same class.
In addition, SymbolicR supports model interpretability through dedicated variable importance metrics, providing insights into the contribution of individual features.
With a median absolute average fold error (AAFE) of 1.18, the extended model shows strong predictive performance and represents a valuable tool for early-stage de-risking in Fc-fusion protein development. The approach is also aligned with the principles of the 3Rs (replacement, reduction, refinement) and the FDA Modernization Act 2.0, supporting more efficient and ethical drug development strategies.
