A Computational Method of Defining Potential Biomarkers based on Differential Sub-Networks

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Network-based analysis of omics data is crucial for biomarker discovery.
• Improving disease diagnosis and early detection of complex diseases requires advanced computational methods.
• Identifying reliable biomarkers is essential for understanding disease mechanisms and progression.

Purpose of the Study:

• To develop and evaluate a computational method, potential biomarkers based on differential sub-networks (PB-DSN), for identifying potential biomarkers.
• To enhance disease diagnosis and detect early warning signals for complex diseases using network analysis.
• To demonstrate the effectiveness of PB-DSN in analyzing both static and time-series omics data.

Main Methods:

• Developed a computational method, PB-DSN, utilizing Pearson correlation coefficient (PCC) to infer networks from feature ratios.
• Extracted differential sub-networks to identify key features for group discrimination and disease emergence.
• Applied topological analysis of differential sub-networks to define potential biomarkers.

Main Results:

• PB-DSN was applied to a static genomics dataset (small, round blue cell tumors) and a time-series metabolomics dataset (hepatocellular carcinoma).
• PB-DSN demonstrated superior performance compared to several established methods including SVM-RFE, empirical Bayes statistics, and graph-based analyses.
• The method effectively identified discriminative features for disease classification and potential warning signals.

Conclusions:

• PB-DSN is an effective computational tool for biomarker discovery through network analysis.
• The method shows significant potential for improving disease diagnosis and identifying early warning signals.
• PB-DSN's performance highlights the value of differential sub-network analysis in omics data interpretation.