ECFS-DEA

ECFS-DEA applies an ensemble classifier-based feature selection strategy to improve differential expression analysis of gene expression profiles by identifying explanatory variables across heterogeneous sample distributions.

Key Features:

• Ensemble Classifier Approach: Integrates multiple base classifiers into an ensemble framework to improve robustness of feature selection.
• Variable Importance Calculation: Implements a variable importance measure inspired by the random forest algorithm that can be adapted to any chosen base classifier.
• Interactive Feature Selection: Supports iterative feature selection via a sorted list of individual variables to refine selected explanatory features.
• Visualization Tools: Generates projection heatmaps using k-means clustering and assesses feature performance with Receiver Operating Characteristic (ROC) curves.
• Multivariate Hypothesis Testing: Applies a multivariate hypothesis testing approach to address limitations of multiple hypothesis testing in detecting collectively explanatory features.

Scientific Applications:

• Differential Expression Analysis: Identifying important variables in differential expression analysis of gene expression profiles.
• Heterogeneous Sample Distributions: Detecting explanatory features across datasets with differing sample distributions.
• Validation and Translational Studies: Applicable from basic research to clinical studies and validated on both simulated and realistic datasets.

Methodology:

Combines an ensemble classifier strategy with a random-forest-inspired variable importance measure and multivariate hypothesis testing, and uses k-means clustering for projection heatmaps and ROC curves for performance assessment.