Goistrat: gene-of-interest-based sample stratification for the evaluation of functional differences

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Understanding gene expression in cancer is vital for personalized medicine.
• Current omics methods struggle to effectively stratify cancer patients for molecular analysis.

Purpose of the Study:

• To develop a novel, biologically-informed workflow for multi-omics data stratification.
• To improve the analysis of cancer patient samples and identify disease mechanisms.

Main Methods:

• Utilized a workflow combining MSigDB gene sets, graph machine learning, and ensemble clustering.
• Applied the method to multi-omics samples with matched RNA-Seq data.
• Validated performance on eight TCGA datasets and analyzed nearly a thousand prostate cancer samples.

Main Results:

• Demonstrated superior sample separation for studying biological differences compared to existing methods.
• Identified specific pathways, including PI3K-AKT-mTOR gene sets, linked to prostate tumor aggressiveness.
• Successfully analyzed the impact of FOLH1 gene expression variation.

Conclusions:

• The developed workflow is a novel tool for identifying disease-relevant gene functions in large datasets.
• Provides a framework for understanding gene expression variation in complex diseases.
• Informs the development of targeted therapeutic strategies.