Exploring Noncovalent Interactions Between Carbohydrates and Proteins: A Computational study

Farzad Mostafavi
MSc Student 
Department of Biochemistry

Date: February 17, 2025
Time: 1:00 p.m. to 2:00 p.m. 
Room: CSF 1302

Abstract:

Interactions between carbohydrates and proteins are crucial in diverse biological processes with implications for human health. Specifically, carbohydrate-protein interactions are essential for cellular recognition and signaling, as well as cell-to-cell adhesion and communication. Dysregulation of these interactions has been proposed to play a role in the development of various diseases such as cancer, neurodegenerative disorders, and metabolic diseases. A comprehensive understanding of carbohydrate-protein interactions is therefore essential for developing targeted therapies. While hydrogen bonding in protein-carbohydrate interactions has been well investigated, hydrophobic interactions—that is, π-interactions involving aromatic amino acids (Tyr, Trp, Phe and His)—have not received as much attention. In this project, we have characterized π-interactions in carbohydrate-protein complexes using computational chemistry and determined the most accurate computational method to compute the interaction energies. We began by mining data from the Protein Data Bank (PDB) to identify the structure of biologically relevant carbohydrate-protein π-interactions. Then, we clustered carbohydrate-protein interactions based on their structural similarity and selected 35 representative interactions. Next, we calculated the interaction energy for each representative using the gold standard method for energy calculations (i.e., CCSD(T)) and 216 different computational methods. In the future, we will use the identified computational method alongside machine learning to calculate the carbohydrate-protein interaction energies based on their structural features. This work will provide a foundation for future studies in drug discovery and development of small-molecule inhibitors that specifically target carbohydrate-mediated biological processes. Additionally, our findings can enhance machine learning models for interaction energy prediction and contribute to improving molecular docking algorithms in computational drug design.