doubletD

doubletD detects doublets in single-cell DNA sequencing (scDNA-seq) data to identify mixed-cell observations that confound analyses of intratumor heterogeneity and tumor evolution.

Key Features:

• Standalone scDNA-seq doublet detection: Focuses exclusively on identifying doublets in scDNA-seq data rather than jointly inferring downstream analyses.
• Maximum likelihood model: Uses a maximum likelihood approach as the core statistical model.
• Closed-form solution: Implements a closed-form solution for the likelihood model to enhance computational efficiency and determinism.
• Performance benchmarking: Demonstrated superior performance compared to methods that jointly infer doublets and downstream analyses and to standalone scRNA-seq doublet detection approaches.
• Evaluation on multiple datasets: Validated on both simulated data and real-world scDNA-seq datasets.
• Improves downstream accuracy: Reduces confounding from doublets to enhance the reliability of downstream analyses of genetic heterogeneity and tumor evolution.

Scientific Applications:

• Intratumor heterogeneity analysis: Removes doublets that can distort assessment of clonal composition and heterogeneity within tumors.
• Tumor evolution studies: Improves inference of evolutionary trajectories by eliminating mixed-cell signals from scDNA-seq data.
• Preprocessing for scDNA-seq pipelines: Serves as a dedicated preprocessing step to increase accuracy of downstream copy-number, mutation, and phylogenetic analyses.

Methodology:

Applies a maximum likelihood approach with a closed-form solution to identify doublets in scDNA-seq data.