gene2gauss: A multi-view gaussian gene embedding learner for analyzing transcriptomic networks.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Gene co-expression network analysis reveals biological processes and regulatory mechanisms.
• Network-based methods offer more biologically relevant insights than standard differential analysis.
• Joint analysis of multiple transcriptomic networks can identify robust gene modules and regulatory networks.

Purpose of the Study:

• To introduce gene2gauss, a novel feature learning framework for gene embedding.
• To leverage multi-study transcriptomic data for enhanced gene representation.
• To identify transcription factor (TF) regulons and gene regulatory networks.

Main Methods:

• Gene embedding into multivariate Gaussian distributions.
• Incorporation of long-range gene interaction neighborhoods across multiple transcriptomic studies.
• Application to idiopathic pulmonary fibrosis (IPF) gene co-expression networks.

Main Results:

• Gene2gauss successfully embeds genes as multivariate Gaussian distributions.
• The learned Gaussian features effectively identify known transcription factor regulons in IPF data.
• Comparison with other feature learning methods shows high relevance of gene2gauss features.

Conclusions:

• Gene2gauss provides a powerful framework for analyzing multi-study transcriptomic data.
• The method enhances the discovery of gene regulatory mechanisms and biological insights.
• Gaussian gene embeddings offer a superior approach for transcriptomic data analysis.