Numbat-multiome: inferring copy number variations by combining RNA and chromatin accessibility information from...

Area of Science:

• Genomics
• Cancer Biology
• Computational Biology

Background:

• Aberrant genomic alterations, including copy number variations (CNVs), are crucial in cancer development.
• The Numbat algorithm is established for CNV inference from single-cell RNA sequencing (scRNA-seq).
• Integrating multi-omic data offers a more comprehensive understanding of cancer heterogeneity.

Purpose of the Study:

• To introduce Numbat-multiome, an extension of Numbat for CNV inference from scRNA-seq and single-cell Assay of Transposase [Tn5]-Accessible Chromatin sequencing (scATAC-seq) data.
• To enable separate or integrated analysis of scRNA-seq and scATAC-seq data for CNV detection.
• To provide a unified computational framework for analyzing multi-modal single-cell data in cancer.

Main Methods:

• Developed Numbat-multiome, unifying scRNA-seq and scATAC-seq data via a common genomic coordinate system and binning strategy.
• Evaluated performance in four modes: RNA gene, RNA bin, ATAC bin, and Combined bin.
• Validated using benchmark tumor cohorts (multiple myeloma, Richter's syndrome) against whole-genome sequencing.

Main Results:

• Numbat-multiome demonstrated robust performance across different CNV event types, achieving high precision and recall (median F1>0.9).
• Consistent performance was observed across various sample types and CNV event lengths.
• The tool successfully tracked clonal evolution and identified rare subclones in serial samples.

Conclusions:

• Numbat-multiome effectively infers CNVs from multi-modal single-cell data, enhancing cancer genomic analysis.
• Integration of epigenomic profiles at the subclonal level provides novel insights into cancer progression.
• This tool facilitates a deeper understanding of the genetic and epigenetic drivers of cancer phenotypic shifts.