Structural analysis of SARS-CoV-2 Spike protein variants through graph embedding

Area of Science:

• Computational Biology
• Structural Biology
• Virology

Background:

• Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants require continuous monitoring due to their impact on protein structure.
• Protein Contact Networks (PCNs) offer a graph-based framework to model protein structures, incorporating biological and topological information.
• Graph embedding methods, particularly using Graph Neural Networks (GNNs), can analyze complex network data.

Purpose of the Study:

• To explore the application of GNNs for embedding PCNs of the SARS-CoV-2 Spike protein.
• To analyze the embedded space to distinguish mutated residues from non-mutated ones in different variants.
• To monitor genetic evolution and identify potentially dangerous outcomes of viral mutations.

Main Methods:

• Construction of PCNs for the wild-type and mutated Spike proteins of SARS-CoV-2.
• Application of the GraphSage embedding algorithm for unsupervised learning of PCN representations.
• Analysis of mutation points within the learned embedding space.

Main Results:

• The study successfully generated PCNs for various SARS-CoV-2 Spike protein variants.
• GraphSage embedding revealed distinct characteristics of mutation points in the embedded space.
• The approach demonstrated the ability to differentiate mutated residues based on their network properties.

Conclusions:

• GNN-based embedding of PCNs is a viable method for analyzing protein structure variations in viral variants.
• This approach can aid in the continuous monitoring of viral evolution and the identification of significant mutations.
• The findings contribute to understanding the structural impact of mutations in SARS-CoV-2, supporting public health surveillance.