dRFEtools: dynamic recursive feature elimination for omics.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Technological advancements yield large omics datasets for machine learning.
• High feature-to-observation ratios and limited sample availability pose challenges.
• Traditional methods overlook peripheral genes in biological networks.

Purpose of the Study:

• Introduce dRFEtools for efficient feature selection in omics data.
• Address limitations of computational cost and feature selection scope.
• Enhance interpretability of machine learning models in biological research.

Main Methods:

• Implement dynamic recursive feature elimination (RFE) for reduced computation.
• Extend dynamic RFE to regression algorithms.
• Integrate with scikit-learn for seamless application in omics data analysis.

Main Results:

• dRFEtools achieves high accuracy with reduced computational time compared to standard RFE.
• Identifies feature subsets with predictive power, including peripheral genes.
• Enhances interpretability of omics data analysis through feature selection.

Conclusions:

• dRFEtools offers a powerful and efficient solution for feature selection in large-scale omics data.
• Facilitates the application of machine learning in biological discovery by improving model interpretability.
• Provides a valuable tool for researchers working with complex biological datasets.