A new method for constructing tumor specific gene co-expression networks based on samples with tumor purity...

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Tumor tissue samples are often a mix of tumor and stromal cells, leading to tumor purity heterogeneity (TPH).
• Ignoring TPH in omics studies can result in inaccurate gene co-expression networks, identifying spurious interactions instead of true biological relationships within tumor cells.

Purpose of the Study:

• To develop a computational method, TSNet, for constructing tumor-specific gene/protein co-expression networks.
• To address the challenge of TPH in omics data analysis.
• To improve the accuracy of network inference in heterogeneous tumor samples.

Main Methods:

• TSNet models observed gene expression profiles as a mixture of cell type expressions.
• The method explicitly incorporates tumor purity percentage for each sample.
• TSNet constructs tumor-cell specific co-expression networks.

Main Results:

• TSNet significantly outperforms standard graphical models that do not account for TPH in synthetic data experiments.
• Application of TSNet to TCGA ovarian cancer RNA-seq data identified novel tumor-specific co-expression modules.
• The analysis revealed a tumor-specific hub structure.

Conclusions:

• TSNet effectively corrects for tumor purity heterogeneity in gene co-expression network construction.
• The method enhances the biological interpretability of omics data from solid tumors.
• TSNet provides a robust approach for identifying tumor-specific molecular interactions.