Genotype-free demultiplexing of pooled single-cell RNA-seq

Abstract

A variety of experimental and computational methods have been developed to demultiplex samples from different individuals mixed in a single-cell RNA sequencing (scRNA-seq) experiment. However, these methods all require extra information is either added to samples (such as sample barcode) or measured from samples prior to mixing (such as genome-wide genotypes). We introduce an alternative approach, in which genetic differences between mixed samples are inferred directly from scRNAseq data without extra information and these differences are used to assign single cells to samples. Our method also identifies a minimal set of presence/absence genotypes which can be used to map samples to their biological source, or to track samples between experiments. We tested our tool "scSplit" on different real and simulated datasets and achieved high true positive rate (> 90%) in mapping cells back to their original source. Doublets can also be sensitively detected as an independent group. Our method is ideally suited to samples for which external genome-wide genotype data cannot be obtained (for example for non-model organisms), or for which it is impossible to obtain unmixed samples directly, such as mixtures of genetically distinct tumour cells, or mixed infections.

Multiplexed single-cell profiling of post-perturbation transcriptional responses to define cancer vulnerabilities and therapeutic mechanism of action

“…To demultiplex the data, as well as to identify droplets containing ambient mRNA (empty droplets) or two cells ('doublets'), we developed a computational method based on SNP-fingerprinting, which classifies single cells with negligible error rates, even for cells with low sequencing depth. During the course of this project, several other methods for SNP-based demultiplexing have been published 20,47,48 . In particular, Demuxlet applies a similar approach, using pre-computed reference SNP profiles for the samples being pooled to identify single cells and detect doublets.…”

Scrublet: Computational Identification of Cell Doublets in Single-Cell Transcriptomic Data

“…d Measures multiplet rate but does not facilitate detection of multiplet cell states in typical experimental samples from a single organism Natural genetic variation (Kang et al, 2018;Xu et al, 2019) By mixing together cells from comparable samples from multiple genotyped individuals, genetic variants in transcripts can be used to assign each cell barcode to one individual, or in the case of multiplets, to multiple individuals. Only inter-individual multiplets can be identified, so the fraction of detectable multiplets increases with the number of individuals.…”

Single-cell transcriptomics reveals that glial cells integrate homeostatic and circadian processes to drive sleep-wake cycles

“…These sample points can be 1) grouped by four Zeitgeber (ZT) times to analyse the transcriptional correlates of circadian rhythms; 2) grouped by "sleep" and "wakefulness" states according to the animal's vigilance status at the time of sample collection to examine "sleep/wakefulness" correlates; and 3) selected and ordered by the fly's level of sleep drive to determine molecular changes associated with the sleep homeostat. To minimize technical batch effects that may mask true biological responses, we applied demultiplexing based on natural variation between wild-type genotypes 13,14 . Specifically, instead of associating each batch to a different sleep or wakefulness state, we associated a different Drosophila Genetic Reference Panel (DGRP) 15 line to each behavioural condition (Fig.…”

“…DGRP lines can be distinguished from one another by their unique SNPs. Their natural genetic variation from each other allows us to determine the condition for each cell from the sequenced data [13][14][15] . This strategy has the added advantage of removing around 90% of droplets that enclose two cells instead of one from the dataset.…”