Identification of gene signatures from RNA-seq data using Pareto-optimal cluster algorithm.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Molecular Diagnostics

Background:

• Gene signatures are crucial for understanding disease mechanisms and classifying samples.
• Existing methods often struggle with complex data and identifying the most informative gene signatures.
• There is a need for powerful, data-driven approaches to detect robust gene signatures.

Purpose of the Study:

• To develop and validate a novel framework for identifying gene signatures from RNA-seq data.
• To utilize multi-objective optimization for robust cluster size identification.
• To establish a gene signature capable of accurate disease classification.

Main Methods:

• A framework combining pre-filtering, normalization, and differential gene expression analysis (Limma) was employed.
• Multi-objective optimization for collecting cluster alternatives (MOCCA) identified Pareto-optimal cluster sizes.
• K-means clustering and Spearman's Correlation Score were used to determine the best gene signature cluster.

Main Results:

• The framework was applied to cervical cancer RNA-seq data, identifying 582 differentially expressed genes (DEGs).
• MOCCA yielded seven Pareto-optimal clusters, with the best cluster containing 35 upregulated DEGs.
• The identified gene signature achieved high classification performance (accuracy 0.935) using PAMR.

Conclusions:

• The developed framework successfully identifies a multi-objective optimized gene signature for disease classification.
• This signature demonstrates high accuracy in distinguishing disease and control samples.
• The method is broadly applicable to RNA-seq and microarray data for signature discovery.