Identifying Structural Variants from Single-Cell DNA Sequencing Data

Abstract

Recent single-cell DNA sequencing technologies enable the sequencing of thousands of cells at single cell resolution, providing an unprecedented opportunity to study intra-tumor genetic heterogeneity. One tradeoff of these technologies is that the data collected per single cell is extremely sparse (0.02 - 0.05X coverage). As a result, recent studies have focused on the analysis of copy-number aberrations (CNAs) in single-cell data. However, CNAs are only a subclass of a larger class of mutations known as structural variants, which have been found to have a significant contribution to human genetic variation and disease. Structural variants (SVs) can affect clonal composition in ways that would be undetectable from an analysis of CNAs; therefore the study of SVs in single-cell data can provide important insights about tumor evolution. No study has yet focused on the identification and analysis of SVs from ultra-low coverage single-cell DNA sequencing data.

We develop a pipeline to identify structural variants from sparse single-cell sequencing data by using the approach of pooling cells together into a pseudo-bulk sample, calling SVs, and then reassigning SVs to individual cells. We then test our pipeline on three different single-cell datasets sequenced on the 10X Genomics Chromium platform and evaluate the validity of the detected SVs. We find that although the detected SVs are too sparse to recover the clonal composition for any of the datasets, we are still able to identify individual SVs that explain previously reported copy number events in the same datasets. Through simulated experiments we also conclude that as single-cell barcoding technologies improve to have higher fidelity barcoding and higher coverage, it will become more feasible to use SVs detected by this pipeline to infer clonal structure.