Increasing coverage from 10x processing

Increasing mito reads / uniform coverage

Caleb Lareau

Overview

We consistently observe low-coverage regions in certain mtDNA regions due to homology with nuclear DNA. In the setting of scATAC-seq, we expect to observe significantly more reads truly generating from the mtDNA than the corresponding nuclear segment. Thus, the idea is to hard-mask regions of the nuclear genome that share consistent homology with the mtDNA, which in effect forces observed reads to the mtDNA when aligned.

Generating hard-mask bed file

Several common reference genomes have been provided already, but a workflow for how to generate your own is discussed in the mitoblacklist repo. The pre-existing blacklist .bed files can be easily downloaded from there.

Updating CellRanger to have a hard-masked nuclear genome

One can follow this code using the reference data from CellRanger 1.0.0 to produce a hard-mask modified reference genome.

One exception is that I couldn't figure out how to make the .flat and .gdx files, nor could I figure out where they were important in the 10X protocol, but the pipeline evidently does need them. So, I just keep them from the old version.

mv genome.fa old_genome.fa
mv genome.fa.flat old_genome.fa.flat
mv genome.fa.gdx old_genome.fa.gdx
bedtools maskfasta -fi old_genome.fa -bed mito_blacklist.bed  -fo genome.fa

# Cleanup old files
rm old_genome.fa 
rm genome.fa.*

# Make new reference genomes
bwa index genome.fa
samtools faidx genome.fa

# Put the flat files back; these won't be hard-masked, but they aren't being used by the aligner
mv old_genome.fa.flat genome.fa.flat
mv old_genome.fa.gdx genome.fa.gdx

# Update the timestamp to pass the CellRanger preflight checks
touch -d "now" genome.fa.flat
touch -d "now" genome.fa.gdx

Update: in version 1.2.0+ of CellRanger-ATAC, the ability to make a full custom reference genome is possible. Thus, a more natural way of generating this blacklist-compatible reference is to 1) generate the hard-masked .fa file and then 2) execute the cellranger-atac mkref command as suggested.

To do this, format the GRCh38_mask.config with a path to the hard masked .fasta file (from the output of bedtools maskfasta) as shown below:

{
    organism: "human" # same as before
    genome: ["GRCh38"] # same as before
    input_fasta: ["/path/to/hardmasked_GRCh38/assembly_hardmask.fa"] # updated from bedtools maskfasta with masked regions
    input_gtf: ["/path/to/gencode/annotation.gtf"] # same as before
    non_nuclear_contigs: ["chrM"] # same as before
    input_motifs: "/path/to/jaspar/motifs.pfm" # same as before
}

This json formatted file coordinates all key components needed to build a reference. Then, with version 1.2.0+ of CellRanger-ATAC, run:

cellranger-atac mkref --config=GRCh38_mask.config

to generate the full reference path.