MLGCN-Driver: a cancer driver gene identification method based on multi-layer graph convolutional neural network.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Cancer progression is driven by mutations in driver genes.
• Identifying cancer driver genes is a key research area.
• Existing methods often overlook high-order network features.

Purpose of the Study:

• To propose a novel method, MLGCN-Driver, for enhanced cancer driver gene identification.
• To incorporate high-order network features into driver gene prediction.
• To leverage multi-omics and topological network data.

Main Methods:

• Developed MLGCN-Driver, a multi-layer graph convolutional neural network (GCN) model.
• Utilized initial residual connections and identity mappings to learn multi-omics features.
• Employed the node2vec algorithm to extract topological structure features.
• Integrated biological and topological features for driver gene probability calculation.

Main Results:

• MLGCN-Driver effectively learns from biological multi-omics and network topological features.
• Residual connections and identity mappings prevent feature over-smoothing.
• The method calculates driver gene probabilities based on integrated features.

Conclusions:

• MLGCN-Driver demonstrates superior performance in driver gene identification on pan-cancer and specific cancer datasets.
• The method achieves excellent results in Area Under the ROC Curve (AUC) and Area Under the Precision-Recall Curve (AUPRC).
• MLGCN-Driver outperforms state-of-the-art approaches in driver gene identification.