Utilizing Graph Neural Networks for Breast Cancer Prognosis Prediction with High-dimensional Genomic Data.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Accurate breast cancer prognosis is crucial for effective treatment and patient well-being.
• Genomic features are high-dimensional and complex, posing challenges for traditional analysis.
• Existing methods may not fully capture intricate gene interactions relevant to cancer prognosis.

Purpose of the Study:

• To develop a novel computational approach for breast cancer prognosis prediction.
• To identify a robust set of prognostic biomarkers from high-dimensional RNA sequencing data.
• To leverage graph neural networks (GNNs) for enhanced extraction of gene interaction patterns.

Main Methods:

• Utilized a systems biology feature selector for dimension reduction on RNA sequencing (RNA-Seq) data.
• Selected 20 prognostic biomarkers strongly associated with breast cancer prognosis.
• Developed a graph neural network (GNN) with a multi-layer perception (MLP) readout on gene interaction networks (GINs).

Main Results:

• Successfully identified 20 prognostic biomarkers relevant to breast cancer.
• The GNN model outperformed baseline models in prediction accuracy.
• Achieved a significant improvement, up to 23% in the area under the precision-recall curve (AUPRC).

Conclusions:

• The proposed GNN-based approach effectively extracts complex genomic interactions for improved breast cancer prognosis.
• This method offers a powerful tool for analyzing high-dimensional genomic data in cancer research.
• The identified biomarkers and GNN model hold potential for clinical decision support in breast cancer treatment.