(628d) Predictive Modeling for Glycan Method Analysis and Species Prediction

Glycans or polysaccharides are carbohydrate-based polymers that are critical for storing energy and system regulation [1]. Several glycan species prevalent in a therapeutic antibody, have shown to affect the pharmacodynamic and pharmacokinetic behavior whereas other glycan structural element can involve adverse immune response [2]. Thus, glycans affect the structural integrity and functionality of proteins. This makes glycan analysis a critical component to further understand how the different species affect the structure and function of the antibodies. “Critical quality attributes are physical, chemical, biological or microbiological properties or characteristics that must be within an appropriate limit, range or distribution to ensure the desired product quality, safety and efficacy” [2]. Glycan analysis forms one component of the critical quality attributes.

Glycosylation includes different types of glycans such as N-linked, O-linked, C-mannosylation, phosphate-linked or glypiation. Within the Biologics Analytical Operations group at Gilead Sciences, glycan analysis is currently conducted using High-Performance Liquid Chromatography studies.

For a study sequence, the current process is to first run a blank sample for system stability, then run the reference standard for system suitability and to note all the glycan species expected within a sequence. The next step is to carry out the glycan analysis for all the samples within a sequence. Further, an analyst then studies these results to evaluate the chromatogram results from the study and note or confirm all the glycan species identified within a sample. Different glycan species elute from a HPLC experiment at different times also known as retention times. Experimental evidence suggests that there is a specific time window for each of the different glycan species.

The current study proposes a predictive modeling based approach to glycan method analysis by applying machine learning techniques for glycan classification. Further, we propose a general data set for glycans to answer broader scope questions for glycan time windows, methods, column health and sample data to identify features of interest to improve efficiency and reduce manual workload. The proposal involves a continuous improvement (CI/CD) pipeline for setting up a continuous data pipeline for glycan analysis.

References:

[1] Yang, Dandan, Zijing Zhou, and Lijuan Zhang. "An overview of fungal glycan-based therapeutics." Progress in molecular biology and translational science 163 (2019): 135-163.

[2] Reusch, Dietmar, and Max L. Tejada. "Fc glycans of therapeutic antibodies as critical quality attributes." Glycobiology 25.12 (2015): 1325-1334.