Enhancing Non-Small Cell Lung Cancer Survival Prediction through Multi-Omics Integration Using Graph Attention...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate cancer survival prediction is crucial for patient management and therapeutic decisions.
• Integrating multi-omics data (mRNA, miRNA, DNA methylation) enhances understanding of cancer's molecular underpinnings.
• Non-small cell lung cancer (NSCLC) survival prediction remains a critical challenge in oncology.

Purpose of the Study:

• To develop and evaluate a novel graph attention network (GAT) model for predicting NSCLC survival.
• To assess the efficacy of integrating multi-omics data for improved survival prediction.
• To identify key molecular features and pathways associated with NSCLC survival.

Main Methods:

• Acquired and preprocessed multi-omics data (mRNA, miRNA, DNA methylation) from The Cancer Genome Atlas (TCGA).
• Utilized chi-square tests for feature selection and SMOTE for dataset balancing.
• Employed a GAT model and measured performance using the concordance index (C-index).

Main Results:

• The GAT model integrating mRNA and miRNA data achieved the highest C-index, demonstrating superior predictive performance.
• Pathway analysis (KEGG) revealed that high-weight features are linked to viral entry pathways (Epstein-Barr virus, Influenza A), implicated in lung cancer.
• The proposed GAT model outperformed existing state-of-the-art methods for NSCLC prediction.

Conclusions:

• A novel GAT-based model effectively predicts NSCLC survival using multi-omics data.
• The integration of mRNA and miRNA data significantly enhances predictive accuracy.
• Identified biological pathways, including viral infections, are critically involved in NSCLC progression and survival.