Assembled graph neural network using graph transformer with edges for protein model quality assessment.

Area of Science:

• Computational biology
• Structural biology
• Bioinformatics

Background:

• Protein structure determination is crucial for understanding protein function.
• Computational methods, particularly deep learning, have advanced protein structure prediction.
• Accurate protein model quality assessment is essential for selecting reliable structures and guiding refinement.

Purpose of the Study:

• To develop an advanced computational method for protein model quality assessment.
• To improve the accuracy of selecting the best protein structure models from predictions.

Main Methods:

• Constructed a graph neural network integrating a Graph Transformer Feature Extractor and message-passing layers.
• Employed a graph-based representation to naturally capture protein structural information.
• Introduced Graph Transformer to weigh the dependencies of neighboring residue nodes.

Main Results:

• The graph-based approach effectively captures spatial patterns and residue dependencies.
• The developed network demonstrated superior performance compared to existing protein quality assessment methods.
• The model is lightweight, utilizing fewer input features and network layers.

Conclusions:

• The proposed Graph Transformer-based graph neural network offers a powerful and efficient solution for protein model quality assessment.
• This method enhances the reliability of computational protein structure prediction.
• The freely available code facilitates further research in structural biology.