switchBox

switchBox provides a framework for reproducible development and evaluation of genomic classifiers from gene expression data to improve the statistical robustness of prognostic signatures.

Key Features:

• Reproducibility: Provides a framework for conducting reproducible analyses, exemplified by re-evaluation of MammaPrint, an FDA-cleared prognostic test for breast cancer.
• Statistical Robustness: Incorporates rigorous statistical methodologies to enhance the reliability of genomic signatures.
• Simplified Classifier Development: Enables derivation of simplified classifiers, demonstrated by a 16-gene prognostic classifier derived from an original 70-gene signature that predicts 5-year disease-free survival with comparable accuracy.
• Algorithmic Approach: Implements pair-wise feature comparisons (TSP) and decision rules such as the majority-wins principle for classifier construction.

Scientific Applications:

• Biomarker and prognostic signature development: Development and evaluation of gene expression–based prognostic classifiers for diseases including breast cancer.
• Personalized medicine research: Supporting identification and validation of genomic classifiers intended for patient stratification and clinical outcome prediction.

Methodology:

Utilizes pair-wise comparisons of features (TSP) with decision rules such as the majority-wins principle.