GTM-decon

GTM-decon infers cell-type-specific gene topic distributions from single-cell RNA sequencing (scRNA-seq) data to deconvolve bulk transcriptomes and characterize cellular composition.

Key Features:

• Automatic Inference: Automatically infers cell-type-specific gene topic distributions from scRNA-seq data.
• Competitive Performance: Demonstrates competitive performance deconvolving simulated and real bulk transcriptome data compared to state-of-the-art methods.
• Sub-cell-type Variations: Infers multiple gene topic distributions per cell type to capture sub-cell-type heterogeneity.
• Phenotype-Specific Distributions: Utilizes phenotype labels from single-cell or bulk data to infer phenotype-specific gene distributions.
• Nested-Guided Design: Employs a nested-guided design to identify cell-type-specific differentially expressed genes from bulk transcriptome data, as shown in bulk breast cancer analyses.

Scientific Applications:

• Oncology: Deconvolves bulk cancer transcriptomes, including breast cancer, to identify cell-type-specific expression patterns and differentially expressed genes.
• Immunology: Dissects immune cell compositions and cell-type-specific expression within complex tissues.
• Biomarker Discovery: Identifies cell-type-specific gene signatures and potential therapeutic targets from bulk and single-cell datasets.

Methodology:

Uses a Guided Topic Model for deconvolution with a nested-guided design to infer cell-type-specific gene topic distributions from scRNA-seq, to infer phenotype-specific gene distributions using phenotype labels from single-cell or bulk data, and to identify cell-type-specific differentially expressed genes from bulk transcriptomes.