cellranger-arc-multiome
Cell Ranger ARC processes 10x Single Cell Multiome ATAC+GEX via mkfastq, count, aggr, and reanalyze, producing paired ATAC fragments and gene expression from the same cells. Use for primary processing
SKILL.mdskills/cellranger-arc-multiome/SKILL.md
GitHubCell Ranger ARC - 10x Multiome ATAC+GEX Pipeline
Purpose
Processes Chromium Epi Multiome data containing paired ATAC and Gene Expression from the same cells. CRITICAL: Cannot analyze ATAC or GEX alone - both modalities required for all pipelines.
Core Workflow
mkfastq: Demultiplex BCL → FASTQ (separate runs for ATAC/GEX flow cells)count: Align, quantify, call cells, perform joint analysis (requires both modalities)aggr: Merge multiple samples with normalizationreanalyze: Re-run secondary analysis with different parameters
Critical Index Structure Differences
ATAC Libraries (Single-Indexed):
- Each sample index = 4 separate oligo sequences (e.g., SI-NA-A1 → 4 oligos)
- bcl2fastq creates 4 separate sample folders per biological sample
- Must merge all 4 in libraries CSV or get 1/4th expected reads
- Read structure: I1 (opt) / R1 / R2 (i5 index) / R3 OR I1 / R1 / I2 (index) / R2
GEX Libraries (Dual-Indexed):
- Single sample index = unique i7 + i5 pair (e.g., SI-TT-A1)
- Standard dual-index structure: I1 (i7) / I2 (i5) / R1 / R2
Libraries CSV Format
fastqs,sample,library_type
/path/to/gex/fastqs,sample_name,Gene Expression
/path/to/atac/fastqs,sample_name,Chromatin Accessibility
Critical Rules:
library_typeis case-sensitive: exactlyGene ExpressionorChromatin Accessibility- Both modalities required (cannot omit either row)
- For split ATAC oligos, add 4 rows with same library_type, different sample names (SI-NA-A1_1, _2, _3, _4)
FASTQ Naming Convention
Must follow bcl2fastq standard: [Sample]_S[N]_L00[Lane]_[Read]_001.fastq.gz
GEX reads: R1 (cDNA), R2 (UMI+barcode), I1/I2 (optional indices) ATAC reads: R1 (genomic), R2 or I2 (barcode), R3 or R2 (genomic), I1 (optional)
Files not matching this pattern will fail - rename before running count.
BCL Convert Critical Parameters
For ATAC libraries, must specify:
OverrideCycles,Y50;I8;U24;Y49 # Outputs UMI to FASTQ
CreateFastqForIndexReads,1 # Required for ATAC
TrimUMI,0 # Critical: preserve UMI
Never include adapter trimming settings (Adapter, AdapterRead1, AdapterRead2 in [Settings]) - damages barcodes/UMIs causing pipeline failure.
Common Pitfalls
ATAC 4-oligo problem: If mkfastq/bcl2fastq created folders like SI-NA-A1_1, SI-NA-A1_2, SI-NA-A1_3, SI-NA-A1_4, you MUST list all 4 in libraries CSV:
fastqs,sample,library_type
/path,SI-NA-A1_1,Chromatin Accessibility
/path,SI-NA-A1_2,Chromatin Accessibility
/path,SI-NA-A1_3,Chromatin Accessibility
/path,SI-NA-A1_4,Chromatin Accessibility
Omitting any = 25% data loss per missing oligo.
Cannot process modalities separately: Running count with only GEX or only ATAC will fail. Use Cell Ranger (GEX) or Cell Ranger ATAC (ATAC) for single-modality analysis.
Cell calling overrides: --min-atac-count and --min-gex-count must be specified together (both or neither). Review default web_summary first before overriding.
Key Output Files
filtered_feature_bc_matrix.h5: Concatenated gene+peak counts (cells only)atac_fragments.tsv.gz: Per-fragment info (critical for downstream peak analysis)analysis/feature_linkage/: Peak-gene regulatory linkagescloupe.cloupe: Loupe Browser file (joint ATAC+GEX visualization)- Separate BAMs:
atac_possorted_bam.bam,gex_possorted_bam.bam
Integration Notes
For downstream analysis:
- Use
filtered_feature_bc_matrix.h5for Seurat/Signac (contains both modalities) - ArchR/SnapATAC2 can import
atac_fragments.tsv.gzdirectly - Peak-gene linkages in
feature_linkage.bedpeinform regulatory network inference - Feature matrix rows = genes + peaks concatenated (check
features.tsv.gzfor boundaries)
Installation: Save to /mnt/skills/user/cellranger_arc.md
Test: cellranger-arc count --help | grep libraries
Token count: ~580 tokens
When not to use
- Do not use for scRNA-only 3'/5' data. Use Cell Ranger (non-ARC) instead.
- Do not use for scATAC-only libraries. Use Cell Ranger ATAC.
- Cannot analyze ATAC or GEX modality alone; both modalities are required for all ARC pipelines.
See also
scvi-multiviscglue-unpaired-multiomics-integrationiterative-peak-mergingpycistopic-atac-topic-modeling
Upstream docs: https://www.10xgenomics.com/support/software/cell-ranger-arc