(164af) Combining the Use of Predictive Modelling Tools and Experimental Data to Optimize Formulation Design for Novel Modalities in Biologics

The recent diversification of biologic molecules in Discovery and Development in the pharmaceutical industry showcases the exciting possibilities of new therapeutics. While protein engineering has enabled the synthesis of complex structures that can be tailored to fit the necessary efficacy, safety, shelf-life, and physicochemical profiles, they also pose new challenges for the developability risk assessment of these constructs utilizing workflows established for more mature modalities (e.g., mAbs). While the decision to progress a molecule through Development involves considering technical and commercial challenges, a key aspect of this decision resides in rapidly and accurately assessing critical liabilities that might present downstream and having in place mitigation strategies to minimize their impact.

The use of predictive computational modeling tools to identify critical quality attributes, mimic stress conditions, and optimize experimental design in early and late Discovery, is one approach that can help teams make data-driven decisions in biologic programs with novel modalities and minimize resources. However, these approaches also need to be adapted to accurately simulate the behavior of these constructs. Therefore, before they can be implemented in standard workflows, they must be benchmarked against appropriate experimental information.

In this work, MOE (Molecular Operating Environment) was used to predict potential liabilities in a fusion protein and compared with experimental results. This novel methodology required that each of the construct subunits were simulated separately. To achieve this, a homology model was built and molecular attributes, post-translational modifications (PTM) and intermolecular interactions were evaluated under several buffer and stress conditions. The results of these computational studies were then used for comparison with experimental findings. The data package included outputs from several characterization techniques that evaluated aggregation, PTM generation, colloidal, physical, and chemical stability. This comprehensive analysis allowed the benchmarking of identified molecule liabilities, highlighted trends, and confirmed the accuracy and value of the predictive modeling. The use of these modeling tools for assessing drug candidates based on novel modalities will optimize the selection of experimental conditions and improve resource allocation to swiftly identify critical liabilities and enable molecule progression.