Deep learning-based prediction of protein-carbohydrate interactions

Project Summary

Protein-carbohydrate interactions govern a wide variety of biological processes and play an essential role in the development of different diseases. In the framework of our project supported by diiP, we have developed DIONYSUS (https://www.dsimb.inserm.fr/DIONYSUS/) [1], the first database of protein-carbohydrate interfaces annotated according to structural, chemical and functional properties of both proteins and carbohydrates. Exhaustive annotation and analysis of the retrieved data allowed us to evaluate the diversity of the currently resolved protein-carbohydrate interfaces and to suggest annotations for previously unannotated proteins. Finally, the obtained data was used to develop a deep learning model for carbohydrate binding site prediction directly from protein sequence using protein language model embeddings.

[1] Gheeraert A, … Galochkina T. DIONYSUS: a database of protein-carbohydrate interfaces. Nucleic Acids Res 2024, DOI: 10.1093/nar/gkae890

Tatiana Galochkina
tatiana.galochkina@u-paris.fr

• Professor (Maître de conférences, HDR) at Université Paris Cité
• laboratory BIGR (INSERM & Université Paris Cité), team DSIMB

Aria Gheeraert