(191ac) Advanced Characterization and Machine Learning for Preclinical Differentiation and Optimization of Lyophilized Vaccine Formulations

The thermal stability and sensitivity of live virus vaccines (LVVs) necessitates a cold supply chain, increasing the cost and logistical challenges associated with transportation and storage of the immunogenic material, limiting the access and scope of immunization coverage globally. One way to improve stability is by formulating these products in the solid state, such as through freeze drying via a lyophilization (LYO) process. Lyophilization has the potential to slow degradation and improve the stability of live-virus vaccines, eliminate cold supply handling, maximize the access communities have to vaccine formulations, and more effectively reduce the cost, morbidity, and mortality of preventable diseases. Traditional approaches of optimizing process parameters may be too conservative, and a need remains to develop alternative product-specific indicators to facilitate process optimization^1,2 and build product knowledge on stability. The work describes efforts to identify and develop novel and effective characterization methods focused on optimizing lyophilized live virus vaccine dosage forms for antigen stability Here, we demonstrate the value of advanced characterization techniques in providing critical insight on stability changes in lyophilized vaccine formulations and driving formulation improvements in preclinical development.

Three placebo vaccine formulations were prepared via lyophilization and staged for stability at the following conditions: -70C, 37⁰C for seven days, and 25⁰C for 45 days. Select methods were executed and systematically assessed for their ability to characterize surface and bulk and analyze for thermal and physical stability. X-Ray Diffraction (XRD) was applied to monitor for phase changes in the solid state, monitoring for crystallinity, or changes in amorphous content, across stressed conditions. The lyophilized materials were analyzed using modulated differential scanning calorimetry (mDSC) to evaluate thermal events and phase changes to amorphous content on stability. The exterior structural features of each formulated LYO cake were imaged by Scanning Electron Microscopy (SEM). The interior attributes and microstructure were analyzed by X-Ray Computed Tomography (XRCT), and machine learning was used to label and segment the reconstructed images to quantify and compare cake attributes.

Our evaluation and development of novel characterization techniques for lyophilized live-virus vaccines yielded promising qualitative and quantitative tools for process and product optimization. These experiments led to an improved understanding of our lyophilized products on stability, where analytical capabilities had been previously limited. XRD and mDSC profiles and SEM and XRCT imaging to monitor qualitative changes on stability across the varied conditions. The methods developed for quantitative image analysis of the 2- and 3-D reconstructed XRCT images were successful in further characterizing and comparing changes in these attributes across each condition on stability. The applied image analysis tools were evaluated for their ability to differentiate interior structures and compositions and establish these machine learning methods as a useful tool for characterizing future lyophilized formulations.

The use of advanced characterization in imaging and machine learning enables vaccine formulators to investigate differences between lyophilized formulations across stressed conditions. The combined use of imaging techniques coupled with and thermal analysis provide a unique perspective into characteristics and bulk properties of lyophilized vaccines that may aid in optimizing stability and performance.

1. Najarian, Jeff et al. “Optimizing lyophilization primary drying: A vaccine case study with experimental and modeling techniques.” International journal of pharmaceutics 659 (2024): 124168. doi:10.1016/j.ijpharm.2024.124168
2. Pu, Yu Elaine et al. “Understanding the Impact of Microstructures on Reconstitution and Drying Kinetics of Lyophilized Cake Using X-ray Microscopy and Image-Based Simulation.” Journal of pharmaceutical sciences vol. 112,6 (2023): 1625-1634. doi:10.1016/j.xphs.2023.01.002