UNDO

UNDO performs unsupervised deconvolution of mixed gene expression data to separate tumor and stromal cell-specific expression profiles and estimate cellular proportions.

Key Features:

• Unsupervised deconvolution: Separates mixed gene expression data into constituent cell-type signals without prior reference profiles.
• Detection of marker genes (MGs): Automatically identifies cell-specific marker genes located on the scatter radii of mixed gene expression data.
• Estimation of cellular proportions: Computes the relative proportions of different cell types in each mixed sample.
• Reconstruction of cell-specific profiles: Deconvolutes mixed expressions to generate cell-type-specific gene expression profiles.
• Data type compatibility: Applicable to microarray gene expression data and adaptable to other quantitative molecular profiling datasets.

Scientific Applications:

• Analysis of heterogeneous tumor samples: Dissects cellular composition in samples containing multiple cell types such as tumor and stroma.
• Tumor purity estimation: Estimates tumor purity in bulk gene expression datasets, including large-scale collections such as TCGA and CPTAC.
• Validation across mixing proportions: Demonstrated performance across a wide range of tumor–stroma mixing proportions on biologically mixed benchmark datasets.

Methodology:

Unsupervised mathematical deconvolution that identifies marker genes on scatter radii of mixed gene expression, estimates cellular proportions, and reconstructs cell-specific expression profiles from bulk quantitative molecular profiling data.