Single cell genetic demultiplexing in the mouse

In this repository we show how to carry out genetic demultiplexing of pooled mouse single cell/nucleus transcriptomics data. We use the software demuxlet (Kang et al. 2018) which can be found at this link. These methods should also work for single nucleus ATAC-Seq as well (see this link).

Below, we show how to run demuxlet in the case where you have heterozygous (possibly outbred) mice, such as Diversity Outbred mice, with genotypes available. These genotypes are typically obtained using genotyping arrays such as the GigaMUGA platform. One can also run demuxlet without genotypes (i.e. freemuxlet) however we do generally recommend genotyping Diversity Outbred mice, so in this repository we focus on the setting where genotypes are available. For inbred mouse strains you should be able to use genetic variation cataloged by the Mouse Genomes Project to construct a VCF file (for example, using bcftools to subset the Mouse Genomes Project VCF) and use this to run demuxlet in an analogous fashion starting on step 2 below.

In all instances, we use the output of 10X CellRanger, specifically the filtered feature-barcode matrix. We then use demuxlet to deconvolute a strain ancestry for each cell, and to identify mixed background doublets.

Singularity containers

For reproducibility, we have produced singularity containers that contain all of the software used throughout this pipeline. You can find links to these containers here:

• demuxlet/popscle
• picard
• R4.3 with tidyverse, qtl2, and other packages
• samtools

Overview of process

1. Construct VCF file for samples in pool
2. Reorder 10X BAM file to play well with VCF in step 1
3. Run dsc-pileup to generate pileups around known variants
4. Run demuxlet to deconvolute strain identity of cells

Construct pooled sample VCF

See the document impute_full_vcf.Rmd. In this document we show how to take inferred allele probabilities from a set of Diversity Outbred mice, and produce a VCF file that we can use for genetic demultiplexing.

This file should then be bgzipped. For example:

 singularity exec containers/samtools_1.10.sif /opt/bin/bgzip impute_full.vcf

Reorder 10X BAM file to play well with VCF

In this repository we include a toy BAM file that consists of two cells that have been extracted from a real 10X RNA-Seq dataset using the utility subset-bam. In a real setting you should use the BAM file produced by 10X CellRanger, called possorted_genome_bam.bam in current (May 2023) versions of CellRanger.

• reorder_bam.slurm - slurm job submission script for doing this reordering on a cluster. If you have large (or many) BAM files (typical), you will want to use a cluster. This file can also be used as a plain bash script.

Run dsc-pileup

You can read more about running demuxlet at this link. In short, we first run dsc-pileup to generate read pileup files, then we run demuxlet to deconvolute pooled samples.

• run_pileup_founders.slurm - slurm job submission script for running dsc-pileup on a cluster. If you have large (or many) BAM files (typical), you will want to use a cluster. This file can also be used as a plain bash script.

Run demuxlet

You can read more about running demuxlet at this link. In short, we first run dsc-pileup to generate read pileup files, then we run demuxlet to deconvolute pooled samples.

• run_demuxlet.slurm - slurm job submission script for running demuxlet on a cluster. This file can also be used as a plain bash script.