Topologically inferring active miRNA-mediated subpathways toward precise cancer classification by directed random...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Conventional gene biomarkers for cancer diagnosis face reproducibility challenges across platforms and patient cohorts.
• Accurate classification biomarkers and disease status are crucial for clinical cancer research.
• MicroRNAs (miRNAs) play significant roles in gene regulation and disease progression.

Purpose of the Study:

• To develop a robust and accurate method for cancer classification using miRNA-mediated subpathway activities.
• To improve upon existing pathway-based, miRNA-based, and gene-based classification approaches.
• To identify reliable biomarkers for sample classification and therapeutic strategy selection.

Main Methods:

• Collected 4775 cancer samples from 12 datasets (4636 TCGA, 139 GEO).
• Developed a directed random walk method (miDRW) to detect miRNA-mediated subpathway activities.
• Constructed a global directed pathway network (GDPN) and identified inversely correlated miRNAs and differentially expressed target genes.
• Integrated topological information, differential expression levels, and miRNA-gene target relationships to calculate subpathway activity.

Main Results:

• The miDRW method demonstrated superior robustness and accuracy compared to existing classification methods.
• High-frequency miRNA-mediated subpathways were identified as reliable classifiers.
• The approach effectively integrates multi-omics data for enhanced cancer subtyping.
• The method showed improved performance in classifying samples across diverse datasets.

Conclusions:

• miRNA-mediated subpathway activity analysis using miDRW provides a more reliable approach for cancer classification.
• The identified high-frequency subpathways can guide the selection of effective therapeutic strategies.
• This method enhances biomarker discovery for clinical cancer diagnosis and research.
• The study highlights the potential of integrating pathway networks and miRNA-gene interactions for precision oncology.