Alleloscope: Integrative single cell analysis of allele-specific copy number alterations and chromatin accessibility in cancer

Abstract

Cancer progression is driven by both somatic copy number aberrations (CNAs) and chromatin remodeling, yet little is known about the interplay between these two classes of events in shaping the clonal diversity of cancers. We present Alleloscope, a method for allele-specific copy number estimation that can be applied to single cell DNA and ATAC sequencing data, either separately or in combination. This approach allows for integrative multi-omic analysis of allele-specific copy number and chromatin accessibility on the same cell. On scDNA-seq data from gastric, colorectal, and breast cancer samples, with extensive validation using matched linked-read sequencing, Alleloscope finds pervasive occurrence of highly complex, multi-allelic copy number aberrations, where cells that carry varying allelic configurations adding to the same total copy number co-evolve within a tumor. The contributions of such allele-specific events to intratumor heterogeneity have been under-reported and under-studied due to the lack of methods for their detection. On scATAC-seq from two basal cell carcinoma samples and a gastric cancer cell line, Alleloscope detects multi-allelic copy number events and copy neutral loss-of-heterozygosity, enabling the dissection of the contributions of chromosomal instability and chromatin remodeling in tumor evolution.

Evolutionary tracking of cancer haplotypes at single-cell resolution

“…Notably, these included focal amplifications of oncogenes such as KRAS, CCNE1 and MYC. Meanwhile, we suggest that pervasive parallel copy number events 10,14,29 in general are a consequence of underlying levels of instability rather than positive selection. On the other hand, how rarer events modify tumour cell fitness remains uncertain, and will require integration of single cell data with evolutionary models of genomic instability 30,31 .…”

HATCHet2: clone- and haplotype-specific copy number inference from bulk tumor sequencing data

“…Unlike other aggregate statistics such as the mirrored BAF (mBAF) [32] or the squared log-odds ratio [33], the mhBAF quantifies the frequency of a specific haplotype across samples, which HATCHet2 uses to identify haplotype-specific CNAs and assign them to tumor clones. To compute mhBAF in HATCHet2, we develop an expectation-maximization algorithm to compute a pseudomaximum likelihood estimate for the mhBAF across different bulk tumor samples by extending existing approaches that have been introduced for single-cell analysis [17,69].…”

“…Reference-based phasing methods [58][59][60]90] exploit this structure using a large database of known haplotypes to identify SNPs that are likely to be grouped together. Following the approach in several previous CNA inference methods [17,24,25,28,69], HATCHet2 applies a reference-based phasing algorithm to infer the phase of alleles across whole chromosomes. Since the accuracy of phasing decreases with genetic distance [91], we use the inferred phase to group together SNPs into haplotype blocks with a user-defined maximum length (25 kilobases by default).…”

“…Moreover, this definition of the mhBAF implies that fp ≤ 0.5 in most samples p, except in samples where the less abundant haplotype across all samples is more abundant in a particular sample; such cases correspond to mirrored-subclonal allelic imbalance [41] and indicate mirrored-subclonal CNAs. We simultaneously estimate the SNP phasing ĥ and the mhBAF fp by adapting an EM algorithm originally applied to single-cell haplotype frequency estimation [69], which extends an earlier EM algorithm used to phase single-cell sequencing data via pseudobulk [17].…”

Single cell biology—a Keystone Symposia report

“…Zhang's group has applied Alleloscope to primary breast tumor and primary and metastasized colorectal and gastric tumors, revealing pervasive intratumor heterogeneity, including highly complex multiallelic copy number aberrations differentiated by haplotype ratios that have previously been underappreciated or ignored. 69 Jean Fan from Johns Hopkins University presented work using computational modeling to infer changes in cellular state from spatially resolved transcriptomic imaging data. Fan's group applies multiplexed error-robust FISH (MERFISH), which uses combinatorial labeling, barcoding, and sequential imaging, to profile spatially resolved genome-wide transcriptomes in single cells within fixed cultures and tissues.…”