MADVAR: a lightweight, data-driven tool for automated feature selection in omics data.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• High-throughput omics data presents analysis challenges due to numerous irrelevant features.
• Traditional feature selection methods are often computationally expensive and rely on arbitrary thresholds.

Purpose of the Study:

• To introduce MADVAR, a lightweight R package for automated feature selection in omics data.
• To present two novel data-driven methods, madvar and intersectDistributions, for threshold definition.

Main Methods:

• MADVAR utilizes two data-driven approaches (madvar and intersectDistributions) to define feature selection thresholds based on data's statistical structure.
• The package is implemented in R, ensuring compatibility across major operating systems.

Main Results:

• MADVAR achieves top performance in clustering and classification tasks across diverse omics datasets.
• The methods efficiently filter features without demanding extensive computational resources, overcoming limitations of traditional approaches.

Conclusions:

• MADVAR offers an efficient and data-driven solution for feature selection in omics data analysis.
• The package seamlessly integrates into existing R-based pipelines, enhancing analytical workflows.