Molecular phenotyping using networks, diffusion, and topology: soft tissue sarcoma.

Area of Science:

• Computational Biology
• Bioinformatics
• Data Science

Background:

• Biological datasets, such as gene expression profiles, are often high-dimensional but contain underlying structures.
• Analyzing complex biological networks and identifying patterns within them is crucial for understanding disease mechanisms.

Purpose of the Study:

• To develop and apply an unsupervised data analysis methodology for multivariate biological datasets integrated with network information.
• To identify coherent states or gene expression signatures within sarcoma data from The Cancer Genome Atlas (TCGA) along the TP53 signaling network.

Main Methods:

• The methodology integrates network geometry (Wasserstein distance), global spectral analysis (diffusion maps), and topological data analysis (Mapper algorithm).
• Applied to RNA-Seq gene expression profiles from sarcoma samples, focusing on genes within specific pathways or networks.
• Analysis was performed along the TP53 (p53) signaling network.

Main Results:

• The analysis successfully identified distinct gene expression signatures within sarcoma subtypes.
• These signatures largely corresponded to known histological subtypes such as leiomyosarcoma, dedifferentiated liposarcoma (DDLPS), and synovial sarcoma.
• A novel signature was discovered, characterized by the activation of SERPINE1 and inactivation of TP73.

Conclusions:

• The developed unsupervised methodology effectively discerns biologically relevant states in high-dimensional gene expression data.
• The identified signatures provide insights into sarcoma heterogeneity and can aid in subtype classification.
• The discovery of a new molecular signature highlights the potential for identifying novel therapeutic targets or biomarkers.