GeneSelectML: a comprehensive way of gene selection for RNA-Seq data via machine learning algorithms.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Machine Learning in Medicine

Background:

• Differential gene expression analysis is crucial for disease research.
• Genomic data modeling, considering gene relationships, enhances predictive performance over univariate analysis.
• Methodological variations in gene selection studies necessitate standardized, user-friendly tools.

Purpose of the Study:

• To develop an open-source, interactive web-based tool for RNA-seq gene selection using machine learning algorithms.
• To provide a platform that simultaneously applies multiple algorithms to avoid missing differentially expressed genes (DEGs).
• To integrate pre-processing, graphical visualization, and gene ontology analysis within a single tool.

Main Methods:

• Development of a web-based tool, GeneSelectML, incorporating six distinct machine learning algorithms.
• Implementation of classical pre-processing steps: filtering, normalization, transformation, and univariate analysis.
• Inclusion of graphical outputs: network plots, heatmaps, Venn diagrams, and box-and-whisker plots, alongside gene ontology analysis for mRNA and miRNA DEGs.

Main Results:

• Application of GeneSelectML to Alzheimer's RNA-seq data identified eleven candidate genes suggested by multiple algorithms.
• The microRNA hsa-miR-148a-3p emerged as a potential novel biomarker for Alzheimer's disease diagnosis.
• Validation of the GeneSelectML tool using the Kidney Chromophobe dataset demonstrated its general applicability.

Conclusions:

• GeneSelectML offers a unique, integrated platform for comprehensive gene selection from RNA-seq data.
• The tool's simultaneous use of diverse machine learning algorithms enhances the reliability of DEG identification.
• GeneSelectML facilitates efficient biomarker discovery and disease mechanism investigation through accessible bioinformatics analysis.