Resolving single-cell copy number profiling for large datasets.

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Single-cell DNA sequencing (scDNA-seq) reveals cancer cell genetic heterogeneity.
• High noise and low coverage in scDNA-seq challenge copy number variation (CNV) estimation.
• Existing CNV tools are slow and struggle with large datasets.

Purpose of the Study:

• Introduce SeCNV, an efficient method for profiling copy numbers from scDNA-seq data.
• Address limitations of existing tools in speed and scalability.
• Accurately estimate CNVs despite noise and low coverage.

Main Methods:

• SeCNV utilizes structural entropy for copy number profiling.
• A local Gaussian kernel constructs a depth congruent map (DCM) to capture bin similarities.
• Genome partitioning by minimizing structural entropy from the DCM enables copy number estimation per segment.

Main Results:

• SeCNV demonstrates robust performance with F1-scores > 0.95 for breakpoint detection in simulations.
• Outperforms state-of-the-art methods significantly.
• Processes datasets >50,000 cells in under 4 minutes, unlike other tools (120h limit).
• Successfully applied to breast cancer datasets, identifying subclones and inferring tumor heterogeneity.

Conclusions:

• SeCNV provides an efficient and accurate solution for CNV profiling in scDNA-seq.
• Enables robust characterization of cancer cell genetic heterogeneity.
• Facilitates large-scale cancer genomics studies.