CoGSPro-net:A graph neural network based on protein-protein interaction for classifying lung cancer-relatrd proteins

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Oncology

Background:

• Lung cancer remains a leading cause of cancer-related mortality worldwide.
• Accurate classification of lung cancer proteins is crucial for developing effective diagnostics and therapeutics.
• Existing methods often struggle to capture complex protein interactions and subtle expression patterns.

Purpose of the Study:

• To develop a novel deep learning algorithm for precise classification of lung cancer-related proteins.
• To leverage protein-protein interaction networks and expression data for enhanced predictive accuracy.
• To identify novel protein biomarkers indicative of lung cancer.

Main Methods:

• Proposed CoGSPro, a deep learning model integrating graph neural networks and attention mechanisms.
• Utilized large-scale protein expression datasets for model training and validation.
• Incorporated protein-protein interaction network information to refine feature extraction.

Main Results:

• CoGSPro achieved a classification accuracy of 96.60% for lung cancer proteins.
• The algorithm outperformed existing baseline methods in protein classification tasks.
• Identified several potential novel biomarkers for lung cancer.

Conclusions:

• CoGSPro demonstrates superior performance in classifying lung cancer proteins.
• The identified biomarkers offer promising avenues for early lung cancer detection and targeted therapies.
• This deep learning approach advances the field of computational oncology and biomarker discovery.