Robust and generalizable CNV detection for single-cell sequencing assays

Area of Science:

• Genomics
• Epigenetics
• Computational Biology

Background:

• Copy number variations (CNVs) are key genomic alterations in cancer and genetic disorders.
• Single-cell resolution is crucial for understanding CNV-driven clonal evolution.
• Existing single-cell RNA sequencing (sc-RNA-seq) CNV detection methods lack accuracy in epigenomic single-cell modalities.

Purpose of the Study:

• To develop a robust, unsupervised method for detecting CNVs across multiple single-cell epigenomic modalities.
• To enable accurate CNV detection and analysis of clonal heterogeneity at single-cell resolution.
• To provide a versatile tool for linking CNV dynamics with epigenetic alterations within individual cells.

Main Methods:

• Developed RIDDLER, an unsupervised method utilizing outlier-aware statistical modeling.
• Employed a robust regression framework to model genome-wide read distribution, accounting for assay-specific biases.
• Applied RIDDLER to single-cell ATAC sequencing (sc-ATAC-seq) and single-cell methylation (sc-methylation) data.

Main Results:

• RIDDLER accurately detects single-cell CNVs and dissects clonal heterogeneity.
• The method demonstrates superior accuracy and robustness to data sparsity compared to competing approaches.
• RIDDLER facilitates the dissection of clonal structure, identification of subclonal accessibility peaks, and multimodal integration.

Conclusions:

• RIDDLER is a scalable, generalizable, and accurate multi-modal method for CNV detection in single cells.
• The tool empowers research linking CNV dynamics to epigenetic alterations.
• RIDDLER advances the field of single-cell epigenomics by providing a powerful CNV detection tool.