Haplotype-enhanced inference of somatic copy number profiles from single-cell transcriptomes

Abstract

Genome instability and aberrant alterations of transcriptional programs both play important roles in cancer. However, their relationship and relative contribution to tumor evolution and therapy resistance are not well-understood. Single-cell RNA sequencing (scRNA-seq) has the potential to investigate both genetic and non-genetic sources of tumor heterogeneity in a single assay. Here we present a computational method, Numbat, that integrates haplotype information obtained from population-based phasing with allele and expression signals to enhance detection of CNVs from scRNA-seq data. To resolve tumor clonal architecture, Numbat exploits the evolutionary relationships between subclones to iteratively infer the single-cell copy number profiles and tumor clonal phylogeny. Analyzing 21 tumor samples composed of multiple myeloma, breast, and thyroid cancers, we show that Numbat can accurately reconstruct the tumor copy number profile and precisely identify malignant cells in the tumor microenvironment. We uncover additional subclonal complexity contributed by allele-specific alterations, and identify genetic subpopulations with transcriptional signatures relevant to tumor progression and therapy resistance. We hope that the increased power to characterize genomic aberrations and tumor subclonal phylogenies provided by Numbat will help delineate contributions of genetic and non-genetic mechanisms in cancer.

Clonally resolved single-cell multi-omics identifies routes of cellular differentiation in acute myeloid leukemia

“…However, these measurements are noisy and biased by cellular identity impeding quantitative analysis. Copy number variants (CNVs) can be inferred from scRNAseq data [15][16][17][18] , but are not always present in AML and other leukemias. Our new computational method, CloneTracer, integrates information from SNVs, mtSNVs and infers CNVs (when present) through a statistical model that appropriately accounts for the noise properties of these measurements.…”

Bayesian inference for copy number intra-tumoral heterogeneity from single-cell RNA-sequencing data

“…Another strength of our fully automated model is that it relies on essentially no prior knowledge. While many other tools either require an additional input of normal reference cells (Tirosh et al, 2016), or rely on manual inspection to determine the number of sub-clones (Gao et al, 2022), our approach automatically distinguishes normal from cancer cells within the input sample and suggests an optimal number of clusters. In all our analyses, we have used the same default parameterizations specified in Section 2.5.…”

“…In benchmarking cell clustering, we compare our model against the classic hierarchical clustering (Figure 2B), which is used by various computational tools such as inferCNV (Tirosh et al, 2016), HoneyBADGER (Fan et al, 2018), and Numbat (Gao et al, 2022). One advantage of our model is that it requires only an upper bound of 𝐾 as input instead of the exact 𝐾.…”

Refphase: Multi-sample reference phasing reveals haplotype-specific copy number heterogeneity

“…The long-range haplotypes derived by Refphase augment existing haplotype-specific approaches for copy number calling in single-cell DNA [29] and RNA [30,50] sequencing technologies, and are distinct from haploblocks produced by population-level statistical phasing approaches [25,51–53] or those derived from chromatin structure data [54,55] or long read sequencing [56,57]. Reference-phasing haploblocks are not limited in length by recombination rates, read lengths, or structural constraints, and instead stretch the full length of the evolutionary gain or loss event that gave rise to the AI, frequently a whole chromosome or chromosome arm.…”

“…Statistical phasing utilises large collections of genotypes [24] and local linkage disequilibrium structure to phase SNPs [25,26]. Multiple groups have previously implemented statistical phasing approaches in the context of whole-genome sequencing (WGS) [27], single-sample bulk sequencing studies [27,28] and in single-cell studies, using DNA [29] and RNA [30]. However, while highly accurate locally, statistical phasing accuracy rapidly decreases with increasing genomic distance, limiting the genomic span within which a SNP and the corresponding SCNA can accurately be assigned to its haplotype-of-origin.…”