Prediction of response to anti-cancer drugs becomes robust via network integration of molecular data

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Biomarker discovery from high-throughput data faces challenges due to data volume and cancer heterogeneity.
• Existing network and pathway approaches are not systematically applied for biomarker validation.
• Dimensionality reduction is crucial for analyzing complex biological datasets.

Purpose of the Study:

• To introduce NEAmarker, a novel method for identifying sensitive and robust pathway-level biomarkers.
• To evaluate NEAmarker's performance across different data types and clinical applications.
• To compare NEAmarker against single-gene and alternative enrichment analyses.

Main Methods:

• Utilizing network enrichment analysis (NEA) to transform gene expression data into a lower-dimensional pathway space.
• Correlating pathway scores with phenotype variables (e.g., drug response, patient survival).
• Validating the method using in vitro anti-cancer drug screening data and The Cancer Genome Atlas (TCGA) clinical data.

Main Results:

• NEAmarker demonstrated superior performance in universal applicability across data types and cross-platform integration.
• Consistent biomarker correlates were found between independent drug screens, indicating robustness.
• The method effectively explained differential patient survival, outperforming existing approaches.
• NEAmarker uniquely identified predictors for both in vitro drug response and patient survival.

Conclusions:

• NEAmarker offers a powerful, versatile approach for biomarker discovery and validation at the pathway level.
• The method addresses cancer heterogeneity and data dimensionality challenges effectively.
• NEAmarker enhances the prediction of drug sensitivity and patient outcomes in cancer research.