CCNMF

CCNMF infers joint clonal structures from paired single-cell DNA sequencing (scDNA-Seq) and single-cell RNA sequencing (scRNA-Seq) data to resolve genomic and transcriptomic clones and investigate the relationship between copy-number alterations and gene expression at single-cell resolution.

Key Features:

• Coupled-Clone Non-negative Matrix Factorization: Simultaneously factorizes scDNA copy-number profiles and scRNA gene expression profiles to resolve genomic and transcriptomic clonal architectures.
• Objective function optimization: Optimizes an objective function that maximizes coherence between clone structures inferred from scRNA-Seq and scDNA-Seq.
• Coupling dosage effect: Models the concordance between copy number and expression via a coupling dosage effect that links the two data types.
• Dosage estimation options: Estimates the coupling dosage effect using linear regression with paired RNA and DNA bulk sequencing data (e.g., TCGA) or by employing an uninformative prior.
• Joint clustering of heterogeneous single-cell omics: Clusters and couples heterogeneous scDNA-Seq and scRNA-Seq data derived from the same specimen for cell-resolved clonal analysis.
• Validation on benchmarks and real data: Demonstrated robustness on simulated benchmarks and applied to ovarian cancer cell line mixtures, gastric cancer cell lines, and primary gastric cancers, revealing high correlations between coexisting genomic and transcriptomic clones.

Scientific Applications:

• Clonal resolution in mixed cell lines: Resolves underlying clonal structures in ovarian cancer cell line mixtures using paired scDNA-Seq and scRNA-Seq.
• Gastric cancer analysis: Identifies concordant genomic and transcriptomic clones in gastric cancer cell lines and primary gastric cancers.
• Copy-number–expression investigation: Enables study of how copy-number alterations influence gene expression at single-cell resolution.

Methodology:

Uses Coupled-Clone Non-negative Matrix Factorization and optimization of an objective function to align clone structures from scDNA-Seq and scRNA-Seq, with the coupling dosage effect estimated via linear regression on paired bulk RNA/DNA data (e.g., TCGA) or an uninformative prior.