(152e) Predictive Model for Glycosylation in Monoclonal Antibodies (mAbs) Using a Knowledge Graph-Based Approach.

Glycosylation is the process by which glycans or carbohydrate-based polymers are attached to another molecule. Glycans are critical for storing energy and system regulation [1]. Glycosylation affects the functional efficacy of monoclonal antibodies (mAbs) are one of the main sources of heterogeneity in mAbs. Glycans can exist on a mAb for many reasons such as inherent expression, cell line, cell culture, engineering, substrate, processing, or posttranslational modifications [2]. Glycosylation is structurally diverse and complex posttranslational modification that can affect the behavior of biological molecules. Diverse glycan structures have shown to have effectson the safety, biological activity, and clearance of a molecule.Thus, characterization of glycan species is a critical component in mAb based development to understand how the different species affect the structure and function of the antibodies [3].

Glycosylation includes different types of glycans such as N-linked, O-linked, C-mannosylation, phosphate-linked or glypiation. In biopharmaceutical product development, glycan analysis is conducted using High-Performance LiquidChromatography studies and further extensive characterization using Mass Spectrometry. A HPLC based glycan study sequence includes multiple samples which are processed using a liquid chromatography column. Different glycan species are eluted at different times also known as retention times. Glycan analysis is a time intensive process as scientists must analyze the chromatography results and manually annotate the different glycans based on the observed results.

Previously, a database was created from all the study sequences performed in the past 3 years. A predictive model was developed based off this database to classify and automatically annotate the observed glycans in a chromatogram. The model utilized data such as peak height, width, retention time, plates, and asymmetry to predict the glycan species in a test sample. However, it was assumed that the model is independent of the type of a molecule.

The current approach evaluates this assumption by appending the glycan database with data from monoclonal antibodies. Particularly, this approach evaluates the effect of molecular attributes such as modality, complementarity-determining region(CDR), heavy chain, light chain sequences, molecular weight, isoelectric point (pI), and extinction coefficient on the glycosylation pattern observed in a sample.

A knowledge graph-based approach is proposed to retain all relevant information regarding the monoclonal antibodies and a modified predictive model is developed to further improve the annotation efficiency and reduce manual effort.

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] Del Val, Ioscani Jimenez, Cleo Kontoravdi, and Judit M. Nagy. "Towards the implementation of quality by design to the production of therapeutic monoclonal antibodies with desired glycosylation patterns." Biotechnology progress 26.6 (2010): 1505-1527.

[3] Kaur, Harleen. "Characterization of glycosylation in monoclonal antibodies and its importance in therapeutic antibody development." Critical Reviews in Biotechnology 41.2 (2021): 300-315.