dRFEtools

dRFEtools implements dynamic recursive feature elimination to perform feature selection in large-scale omics datasets, enabling identification of core and peripheral genes and reducing computational time while retaining predictive accuracy.

Key Features:

• Dynamic recursive feature elimination (dRFE): An extension of recursive feature elimination that iteratively removes features to reduce computational time while maintaining predictive accuracy.
• Regression algorithm support: Supports regression algorithms in addition to traditional classification-focused RFE applications.
• Predictive feature subset output: Outputs subsets of features that retain predictive power both with and without peripheral features, facilitating analysis of core and peripheral genes.
• Integration with scikit-learn: Integrates with scikit-learn to leverage its estimators and workflows for model fitting and selection.
• High-dimensional omics applicability: Designed for large-scale omics datasets with high feature-to-observation ratios.
• Python 3 implementation: Implemented in Python 3 for compatibility with Python-based analysis environments.
• Enhanced interpretability: Produces results that support interpretation of selected features and gene-network relationships.

Scientific Applications:

• Feature selection in omics: Selecting informative features from high-dimensional transcriptomic and other omics datasets.
• Core and peripheral gene identification: Identifying core genes and peripheral genes within gene networks that contribute to predictive models.
• Regression-based predictive modeling: Applying regression algorithms for continuous outcome prediction using selected omics features.
• Computational cost reduction: Reducing runtime and resource requirements for feature selection in high-dimensional data.

Methodology:

Implements dynamic recursive feature elimination (dRFE), an iterative extension of recursive feature elimination (RFE) that integrates with scikit-learn, supports regression algorithms, and outputs predictive feature subsets.