Wx: a neural network-based feature selection algorithm for transcriptomic data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Next-generation sequencing (NGS) generates vast amounts of data, posing challenges in selecting optimal gene sets for analysis.
• Identifying reliable gene expression biomarkers is crucial for distinguishing between normal and diseased states, particularly in cancer.

Purpose of the Study:

• To develop a novel algorithm, Wx, for efficient and accurate gene selection from complex genomic datasets.
• To identify a set of robust gene expression biomarker candidates using the Wx algorithm for pan-cancer analysis.

Main Methods:

• Development of a neural network-based feature selection algorithm named Wx.
• Ranking genes based on the discriminative index (DI) score to assess their classification power.
• Application of the Wx algorithm to a TCGA pan-cancer gene-expression cohort.

Main Results:

• The Wx algorithm successfully identified 14 gene-expression biomarker candidates capable of distinguishing cancer from normal samples across 12 cancer types.
• The identified biomarkers demonstrated performance comparable to or exceeding previously reported universal gene expression biomarkers.
• The Wx algorithm proved effective for analyzing next-generation sequencing data in a pan-cancer context.

Conclusions:

• The Wx algorithm offers a valuable tool for selecting optimal gene sets from NGS data, facilitating biomarker discovery.
• Wx provides a complementary approach to existing analytical methods for identifying gene expression biomarkers.
• The identified gene candidates hold potential for improving cancer diagnostics and research.