Extraction of the polysorbate 20 and 80 fingerprint via generative modeling

Area of Science:

• Analytical Chemistry
• Biopharmaceutical Analysis
• Computational Chemistry

Background:

• Polysorbates 20 (PS20) and 80 (PS80) are critical excipients for biopharmaceutical stabilization.
• Their inherent heterogeneity and susceptibility to degradation complicate analytical characterization.
• Current analytical methods are often labor-intensive and lack comprehensive molecular detail.

Purpose of the Study:

• To develop an automated, in-depth analytical method for characterizing complex polysorbate mixtures.
• To enable precise identification of individual polysorbate subspecies and their degradation products.
• To provide a robust tool for quality control and degradation monitoring in biopharmaceutical manufacturing.

Main Methods:

• Development of a hierarchical generative model for liquid chromatography-mass spectrometry (LC-MS) data.
• Integration of domain knowledge including base structures, oxyethylene chain lengths, and esterification.
• Reconstruction of entire LC-MS measurements for automated data interpretation.

Main Results:

• The model successfully resolves individual polysorbate subspecies and determines molecular-level composition.
• Distinguishes between oxidative and hydrolytic degradation pathways, providing pathway-specific fingerprints.
• Achieved high agreement with manual integration, offering enhanced depth and automation.

Conclusions:

• The developed model transforms polysorbate analysis into an objective and comprehensive characterization tool.
• Enables automated, detailed analysis of polysorbate heterogeneity and degradation.
• Supports critical biopharmaceutical applications such as quality control, batch selection, and stability monitoring.