Accurate and fast feature selection workflow for high-dimensional omics data.

Area of Science:

• Bioinformatics and computational biology
• Machine learning in omics data analysis

Background:

• Biomedical research is experiencing a 'Big Data' surge, increasing data volume (D) through more samples (S) and features (F).
• Omics data often contains redundant and irrelevant features, leading to computational inefficiencies and suboptimal model performance.
• Selecting the appropriate feature selection (FS) algorithm for complex bioinformatics problems remains a significant challenge.

Purpose of the Study:

• To develop an efficient workflow and an R package for feature selection in bioinformatics machine learning.
• To address challenges in choosing the correct FS algorithm and strategy for omics data.

Main Methods:

• Selected a subset of feature selection methods for an efficient workflow.
• Developed an R package to implement the feature selection workflow.
• Evaluated the workflow using seven diverse proteomics and gene expression datasets.

Main Results:

• The developed workflow and R package streamline the feature selection process in bioinformatics.
• The evaluation provides guidance for selecting appropriate feature selection strategies.
• Demonstrated the utility of the workflow across different omics datasets.

Conclusions:

• The proposed workflow and R package offer a practical solution for feature selection in bioinformatics machine learning.
• This approach helps mitigate issues related to redundant features and improves model performance.
• Provides a framework to guide researchers in navigating the complexities of feature selection for omics data.