Good practice guidelines for biomarker discovery from array data: a case study for breast cancer prognosis

Area of Science:

• Biomolecular research
• Genomics
• Translational medicine

Background:

• Biomarker discovery is crucial for personalized medicine but faces challenges in clinical validation.
• Few molecular biomarkers from high-dimensional data achieve successful clinical application.
• Concerns exist regarding the reliability of current biomarker discovery methods.

Purpose of the Study:

• To propose good practice guidelines for robust biomarker discovery.
• To introduce a novel tool for identifying and validating biomarkers.
• To apply and demonstrate the approach using breast cancer prognosis data.

Main Methods:

• Development of good practice guidelines for biomarker discovery.
• Implementation of a novel computational tool for feature selection.
• Application to a public breast cancer prognosis dataset.
• Validation using an independent cross-platform dataset.

Main Results:

• Identification of a small set of predictive markers for specific breast cancer patient subpopulations.
• Demonstration of favorable performance compared to existing prognostic tools.
• Validation of high-quality feature selection, with many markers showing individual efficacy.
• Discovery of markers particularly effective for young, estrogen receptor-positive, lymph node-negative early-stage breast cancer patients indicating high recurrence risk.

Conclusions:

• Adherence to good practice guidelines enables the identification of highly predictive genes from high-dimensional breast cancer data.
• The identified predictive genes were successfully validated on an independent, cross-platform dataset.
• The findings suggest potential for improved treatment strategies for specific high-risk patient subsets.