Predicting Risk in Type 2 Diabetes and Clear Cell Renal Cell Carcinoma with Glycosylation Data and Machine Learning

Glycosylation is an essential post-translational modification of proteins that influences many physiological and pathophysiological processes. Altered glycosylation patterns are associated with a variety of human diseases, including type 2 diabetes (T2D) and cancer, and may serve as novel biomarkers for early detection, diagnosis, or prognosis. Recent advances in analytical techniques have made it possible to explore the glycome and glycoproteome for potential biomarkers and therapeutic targets. However, the use of artificial intelligence for extracting clinically relevant insights from glycosylation data remains limited. This study aimed to investigate the predictive potential of glycomic and glycoproteomic data using machine learning (ML) algorithms. A machine learning pipeline was used to develop predictive models from two independent datasets. The first dataset included plasma N-glycan baseline profiles from 74 FinRisk participants, with 37 individuals developing T2D and 37 remaining normoglycemic. The second dataset contained N-glycopeptide abundances from 103 CPTAC clear cell renal cell carcinoma (ccRCC) tumors. The T2D model, an XGBoost classifier, predicted the risk of T2D diagnosis within ten years and achieved an AUC of 0.812. The ccRCC model, a multilayer perceptron classifier, predicted the risk of progressive disease within five years of surgical resection and yielded an AUC of 0.943. The results demonstrate the potential of glycomic and glycoproteomic data for risk prediction in T2D and ccRCC and underscore the value of ML for detecting subtle patterns in complex glycosylation datasets. Furthermore, this study marks the first integration of ML-ready datasets with the GlyGen knowledgebase and glycosylation-related models within the PredictMod platform, enhancing the accessibility and utility of glycosylation data and predictive models for potential clinical and research applications.

Author

• Martinez, Karina

Language

• English

Keyword

• Cancer
• Diabetes
• Artificial Intelligence
• Machine learning
• Biomarker
• Glycosylation

Date created

• 2024

Type of Work

• Master's Thesis

License

• All rights reserved

Rights statement

• In Copyright

GW Unit

• Bioinformatics and Molecular Biochemistry

Degree

• M.S.

Advisor

• Mazumder, Raja

Persistent URL

•  https://scholarspace.library.gwu.edu/etd/q811kk58c