Solo workflow for the 10x multiomics analysis

[akhst7]

I have been trying to create the pipeline to analyze 10x multiome data (PBMC from a Healthy Donor - Granulocytes Removed Through Cell Sorting (3k)) which contains both GEX and ATAC sequencing results from single nuclei. I have a following working pipeline and I'd appreciate any comments and suggestions.

1. GEX and ATAC mapping; I use following Solos scripts for GEX and ATAC respectively ;

GEX:

#!/opt/homebrew/bin/zsh

index=/Volumes/MySlateDrive/star_genome_index/index_human_10x
whitelist=/Volumes/MySlateDrive/10X_Whitelist/737K-arc-v1.txt
r1=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/gex/pbmc_granulocyte_sorted_3k_S1_L003_R1_001.fastq.gz
r2=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/gex/pbmc_granulocyte_sorted_3k_S1_L003_R2_001.fastq.gz
r3=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/gex/pbmc_granulocyte_sorted_3k_S1_L004_R1_001.fastq.gz
r4=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/gex/pbmc_granulocyte_sorted_3k_S1_L004_R2_001.fastq.gz

STAR --genomeDir $index \
--runThreadN 40 \
--readFilesCommand 'gzip -c -d' \
--readFilesIn $r2,$r4 $r1,$r3 \
--soloCBwhitelist $whitelist \
--soloType CB_UMI_Simple \
--soloCBstart 1 \
--soloCBlen 16 \
--soloUMIstart 17 \
--soloUMIlen 12 \
--soloStrand Forward \
--soloBarcodeReadLength 0 \
--soloBarcodeMate 0 \
--clipAdapterType CellRanger4 \
--twopassMode Basic \
--soloMultiMappers EM \
--soloFeatures Gene GeneFull Velocyto \
--outFilterMultimapNmax 10 \
--outFilterScoreMin 30 \
--soloCBmatchWLtype 1MM_multi_Nbase_pseudocounts \
--soloUMIfiltering MultiGeneUMI_CR \
--soloUMIdedup 1MM_CR \
--soloCellFilter  EmptyDrops_CR 5000 0.99 10 45000 90000  500  0.01  20000  0.01  10000 \
--soloCellReadStats Standard \
--outReadsUnmapped Fastx \
--outFileNamePrefix /Volumes/MySlateDrive/pbmc3Ksorted.GEX2 \
--outSAMtype None 

ATAC:

#!/opt/homebrew/bin/zsh

genomedir=/Volumes/MySlateDrive/star_genome_index/index_human_Gencode_GRCh38_p13
tempdir=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac
r1=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L001_R1_001.fastq.gz
r2=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L001_R2_001.fastq.gz
r3=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L001_R3_001.fastq.gz
r4=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L002_R1_001.fastq.gz
r5=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L002_R2_001.fastq.gz
r6=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L002_R3_001.fastq.gz
r7=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L003_R1_001.fastq.gz
r8=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L003_R2_001.fastq.gz
r9=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L003_R3_001.fastq.gz
r10=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L004_R1_001.fastq.gz
r11=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L004_R2_001.fastq.gz
r12=/Volumes/MySlateDrive/pbmc_granulocyte_sorted_3k/atac/pbmc_granulocyte_sorted_3k_S12_L004_R3_001.fastq.gz
whitelist=/Volumes/MySlateDrive/10X_Whitelist/737K-arc-v1.txt
out=/Volumes/MySlateDrive/atac_Solo

STAR  \
        --genomeDir $genomedir \
        --readFilesCommand 'gzip -c -d' \
        --runMode alignReads \
        --runThreadN 40 \
        --readFilesIn $r1,$r4,$r7,$r10 $r3,$r6,$r9,$r12 $r2,$r5,$r8,$r11\
        --soloType CB_samTagOut \
        --soloCBwhitelist $whitelist \
        --soloBarcodeReadLength 0 \
        --sjdbGTFfeatureExon exon \
        --soloCBmatchWLtype 1MM \
        --soloStrand Forward \
        --outSAMattributes NH HI CR CB CY UR UY sM\
        --outSAMtype BAM Unsorted \
        --outReadsUnmapped Fastx \
        --outFileNamePrefix $out

I also run CellRanger ARC (v2.0.2) as follows in my institutional HPC for comparison;

#!/bin/bash

#SBATCH -A general
#SBATCH -o cellranger.arc_%j.txt
#SBATCH -e cellranger.arc_%j.err
#SBATCH --mail-type=ALL
#SBATCH [email protected]
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --nodelist=c92
#SBATCH --cpus-per-task=64
#SBATCH --signal=2
#SBATCH --no-requeue
#SBATCH --time=48:00:00
#SBATCH --mem=128G
#SBATCH --exclusive
#SBATCH --job-name="Cellranger_ARC_pbmc3k.sorted"

#Switch to zsh
zsh

#change directory to slate project
cd /N/slate/

#Load any modules that your program needs
module load openmpi

#Run your program
export OMP_NUM_THREADS=128

srun cellranger-arc count --id=pbmc3ksorted \
                          --reference=/N/slate/ahoji/refdata-cellranger-arc-GRCh38-2020-A-2.0.0 \
                          --libraries /N/slate/ahoji/pbmc_granulocyte_sorted_3k_library.csv \
                          --gex-exclude-introns=true \
                          --localcores=64 \
                          --localmem=128

I am comparing outputs by CellRanger-ARC to those by StarSolo (STAR version=2.7.10b_alpha_23-06-09).

For GEX, there is a good agreement in CB counts between CellRanger ARC and STAR Solo.

CB From CellRanger-ARC output:

 gzcat barcodes.tsv.gz|gsed 's/-1//'|sort|uniq > barcode.sorted
  wc -l barcode.sorted
  2707 barcode.sorted

CB From StarSolo output (in Gene/filtered/):
  wc -l barcode.sorted
  2738 barcode.sorted

Overlapping CBs between those outputs:
  comm -12 barcode.sorted         ../../snatac_snRNAseq_pbmc3k/pbmc3Ksorted.GEX.Solo/Gene/filtered/barcode.sorted |wc -l
   2668

2. ATAC and GEX barcode matching; ATAC and GEX ATAC and GEX uses the different barcodes, they have to be matched in the final output. 10x calls this process as "barcode translation" (https://kb.10xgenomics.com/hc/en-us/articles/360049105612-Barcode-translation-in-Cell-Ranger-ARC). Cell Ranger-ARC does this automatically, however, for Solo outputs, this has to be done manually. Barcode matching simply is the match between line numbers of GEX and those of ATAC barcodes in the ARC-Whitelist.

First thing first is to find corresponding line numbers of GEX bardoes from a barcode.tsv to the ARC GEX barcode whitelist (737K-arc-v1.gex.txt).

#Barcode matching between GEX and ATAC

rg -nf barcodes.tsv ../../../10X_Whitelist/737K-arc-v1.gex.txt -j 40 ---dfa-size-limit 1G > ../../../GEX.barcode.selected.txt

head GEX.barcode.selected.txt
20:AAACAGCCAAATATCC
232:AAACAGCCAGGAACTG
253:AAACAGCCAGGCTTCG
925:AAACCAACACCTGCTC
967:AAACCAACAGATTCAT
1048:AAACCAACAGTTGCGT
1055:AAACCAACATAACGGG
1068:AAACCAACATAGACCC
1709:AAACCGCGTGAGGTAG
2600:AAACGCGCATACCCGG

To get line numbers

gsed 's/:\w\+//' GEX.barcode.selected.txt > GEX.barcode.linenumber.txt

head GEX.barcode.linenumber.txt
20
232
253
925
967
1048
1055
1068
1709
2600

To get cell barcodes based off the line number

awk 'NR == FNR{a[$0]; next};FNR in a' GEX.barcode.linenumber.txt 10X_Whitelist/737K-arc-v1.atac.txt >atac.barcode.selected

head atac.barcode.selected
ACAGCGGGTTGTGACT
ACAGCGGGTCCGTTAG
ACAGCGGGTAACCGGG
CTTTATCGTACACCGC
CTTTATCGTTATTTGG
CTTTATCGTGAACGGT
CTTTATCGTCCTCCCA
CTTTATCGTAAAGGTT
AATAGCTCAGGACCTG
ATAACCCGTCACCATG

3. There are a several tools available to generate fragments and peak files for downstream ATAC analysis; sinto (https://github.com/timoast/sinto) and snapATAC2 for the fragment, and MACS3-hmmratac and genrich for the peak file. Genrich has an internal parameter to remove overlapping fragments from black.list.hg38.bed whears MACS3-hmmatac needs to have additional step to manually remove them. Then, merged peaks between genrich and MACS3-hmmatac peaks were used as an input peak file.

Seurat and Signac were used for further analyses and visualization. Prior to this step, doublets and lower expression cells were removed by scDBlFinder and scttule. Following shows UMAP of single GEX and ATAC assays as well as an integrated GEX and ATAC assays. Results from Cell Ranger outputs are also shown as a comparison.

Solo;

Cell Ranger;

[alexdobin]

Hi @akhst7

This is very interesting; thanks for sharing your pipeline.
One suggestion I have is to use --soloFeatures GeneFull_Ex50pAS for RNA which will make the results closer to CellRanger's.