DualMarker: A Multi-Source Fusion Identification Method for Prognostic Biomarkers of Breast Cancer Based on...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Breast cancer prognosis prediction is challenging due to complex contributing factors.
• Network-based approaches are used to identify prognostic biomarkers, but single-source networks yield poor accuracy.
• Existing multi-source network integration methods often neglect network denoising, limiting performance.

Purpose of the Study:

• To develop an accurate method for identifying prognostic biomarkers in breast cancer.
• To address the limitations of single-source networks and improve multi-source network integration.

Main Methods:

• Proposed DualMarker, a multi-source fusion method for prognostic biomarker identification.
• Constructed a dual-layer heterogeneous network integrating multiple biological data sources.
• Applied network denoising to mitigate effects of incomplete biological network interactions.
• Utilized network propagation and an initial scoring strategy for feature ranking.

Main Results:

• DualMarker demonstrated superior performance across six breast cancer datasets compared to six other network-based methods.
• Identified biomarkers showed biological interpretability.
• The identified biomarkers were closely associated with breast cancer patient prognosis.

Conclusions:

• DualMarker offers an effective approach for identifying reliable prognostic biomarkers for breast cancer.
• The method's integration of multi-source networks and denoising significantly improves prediction accuracy.
• The identified biomarkers hold potential for clinical application in breast cancer outcome prediction.