dragenflow

A Nextflow DSL2 pipeline for running DRAGEN alignment and variant calling workflows. Built on the nf-core framework, it supports multiple analysis modes (germline, somatic, RNA-seq, methylation) with configurable compute profiles for on-premise DRAGEN hardware and AWS.

Table of Contents

• Pipeline Overview
• Analysis Modes
• Samplesheet Formats
• Run Command
• Compute Profiles
• Testing with Example Data
• Key Parameters
• Outputs
• Reference Files

---

Pipeline Overview

All analysis modes share these common steps:

1. Parse samplesheet — validates and routes samples to alignment or demux entry points
2. Gather alignment samples — collects FASTQs from reads, fastq_lists, demux paths, or converts CRAMs/BAMs back to FASTQ
3. Create fastq_list — generates per-sample DRAGEN fastq_list CSV
4. DRAGEN alignment — runs dragen with mode-specific arguments
5. Variant annotation — VEP annotation of SNVs, SVs, and CNVs → TSV tables
6. MultiQC — aggregates QC metrics

Mode-specific downstream steps are described in Analysis Modes below.

---

Analysis Modes

Each mode is activated by passing the corresponding profile. The workflow parameter is set automatically by the profile.

alignonly

Alignment only — no variant calling.

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,alignonly \
  --input samplesheet.csv \
  --outdir results/

germline

Germline SNV, SV, and CNV calling. Outputs VEP-annotated VCFs and TSV tables for each variant type.

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,germline \
  --input samplesheet.csv \
  --outdir results/

DRAGEN flags enabled: --enable-variant-caller, --enable-sv, --enable-cnv

somaticheme

Somatic tumor/normal calling for hematologic malignancies (liquid tumor mode). Requires a samplesheet with matched tumor and normal samples. Includes SNV, SV, and CNV calling with systematic noise filtering.

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,somaticheme \
  --input somatic_samplesheet.csv \
  --outdir results/

Default noise files (hg38):

• SNV: IDPF_WGS_hg38_v.2.0.0_systematic_noise.snv.bed.gz
• SV: WGS_FF_Heme_hg38_v3.1.0_systematic_noise.sv.bedpe.gz

somaticsolid

Somatic tumor/normal calling for solid tumors. Similar to somaticheme but without liquid tumor UMI options; includes TMB estimation.

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,somaticsolid \
  --input somatic_samplesheet.csv \
  --outdir results/

tumoronlyheme

Tumor-only variant calling for hematologic malignancies — no matched normal required. Enables ploidy estimation, DUX4 fusion caller, and CNV calling against a population VCF.

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,tumoronlyheme \
  --input samplesheet.csv \
  --outdir results/

Default population VCF: 1000G_phase1.snps.high_confidence.hg38.vcf.gz

rnaseq

RNA-seq alignment with quantification and gene fusion detection. Downstream steps annotate gene/transcript expression tables.

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,rnaseq \
  --input samplesheet.csv \
  --outdir results/

DRAGEN flags enabled: --enable-rna, --enable-rna-quantification, --enable-rna-gene-fusion, --rrna-filter-enable

Default annotation GTF: Homo_sapiens.GRCh38.105.chr.sorted.gtf.gz

bsseq

Bisulfite sequencing — whole-genome methylation calling (5mC). Downstream steps generate CpG methylation BED and BigWig files.

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,bsseq \
  --input samplesheet.csv \
  --outdir results/

Uses a methylation-specific DRAGEN reference directory (dragen_hg38_5mCv4.3.6).

idtumi

UMI-aware adapter processing using IDT UDI-UMI (10x19) format. Disables duplicate marking (UMI deduplication is used instead). Combine with another analysis profile for variant calling.

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,idtumi,tumoronlyheme \
  --input samplesheet.csv \
  --outdir results/

---

Samplesheet Formats

The pipeline accepts several CSV formats. The first column is always id (or a pair-level ID for wide somatic format).

Standard samplesheet

For most modes (germline, alignonly, rnaseq, bsseq, tumoronly):

id,read1,read2,fastq_list,cram,bam
sample1,/path/to/sample1_R1.fastq.gz,/path/to/sample1_R2.fastq.gz,,,
sample2,,,/path/to/sample2_fastq_list.csv,,
sample3,,,,/path/to/sample3.cram,

Columns are optional — provide whichever input type is available. Multiple input types for the same sample are merged (e.g., existing CRAM + new FASTQs will both be realigned together), as in assets/stub/cram_fastq_wgs_mastersheet.csv, where 10929-BM-lib1 is realigned from its existing CRAM plus an additional lane of FASTQs:

id,cram,read1,read2
10940-BM-lib1,assets/stub/crams/10940-BM-lib1_tumor.cram,,
10929-BM-lib1,assets/stub/crams/10929-BM-lib1_input.cram,,
10929-BM-lib1,,assets/stub/demux_fastq/10929-BM-lib1_S1_L007_R1_001.fastq.gz,assets/stub/demux_fastq/10929-BM-lib1_S1_L007_R2_001.fastq.gz

A BAM can be substituted for a CRAM the same way — see assets/stub/bam_wgs_test_mastersheet.csv:

id,bam
10929-BM-lib1,assets/stub/crams/10929-BM-lib1_input.bam

DRAGEN fastq_list

A pre-built DRAGEN-format fastq_list can be referenced from the fastq_list column, as in assets/stub/fastq_list_wgs_mastersheet.csv:

id,fastq_list
10929-BM-lib1,assets/stub/demux_fastq/fastq_list.csv

where the referenced file is a standard DRAGEN fastq_list (one row per lane):

RGID,RGSM,RGLB,Lane,Read1File,Read2File
TAATGTGTCT.TATGCCTTAC.5,10929-BM-lib1,UnknownLibrary,7,assets/stub/demux_fastq/10929-BM-lib1_S1_L007_R1_001.fastq.gz,assets/stub/demux_fastq/10929-BM-lib1_S1_L007_R2_001.fastq.gz
TAATGTGTCT.TATGCCTTAC.6,10929-BM-lib1,UnknownLibrary,8,assets/stub/demux_fastq/10929-BM-lib1_S1_L008_R1_001.fastq.gz,assets/stub/demux_fastq/10929-BM-lib1_S1_L008_R2_001.fastq.gz

Somatic — long format (recommended)

Tumor and normal samples are listed as separate rows sharing the same pair-level id, distinguished by sample_id and sample_type. See assets/stub/somatic_reads_mastersheet.csv:

id,sample_id,sample_type,read1,read2
KDM7A-KO-CART,CD34-CART-DNA,normal,assets/stub/demux_fastq/CD34-CART-DNA_S6_L003_R1_001.fastq.gz,assets/stub/demux_fastq/CD34-CART-DNA_S6_L003_R2_001.fastq.gz
KDM7A-KO-CART,KDM7A-KO-CART-DNA,tumor,assets/stub/demux_fastq/KDM7A-KO-CART-DNA_S3_L005_R1_001.fastq.gz,assets/stub/demux_fastq/KDM7A-KO-CART-DNA_S3_L005_R2_001.fastq.gz

Somatic — wide format

Tumor and normal can also be specified in a single row using prefixed columns. See assets/stub/somatic_reads_wide_mastersheet.csv:

id,tumor_id,normal_id,tumor_read1,tumor_read2,normal_read1,normal_read2
KDM7A-KO-CART,KDM7A-KO-CART-tumor,KDM7A-KO-CART-normal,assets/stub/demux_fastq/KDM7A-KO-CART-DNA_S3_L005_R1_001.fastq.gz,assets/stub/demux_fastq/KDM7A-KO-CART-DNA_S3_L005_R2_001.fastq.gz,assets/stub/demux_fastq/CD34-CART-DNA_S6_L003_R1_001.fastq.gz,assets/stub/demux_fastq/CD34-CART-DNA_S6_L003_R2_001.fastq.gz

Or referencing a samplemap. See assets/stub/somatic_samplemap_wide_mastersheet.csv:

id,tumor_id,normal_id,samplemap
KDM7A-KO-CART,KDM7A-KO-CART-DNA,CD34-CART-DNA,assets/stub/demux_fastq/Samplemap2.somatic.csv

MGI Samplemap format

MGI sequencer output (Samplemap2.csv) is also accepted directly via the samplemap column. The Library Name field in the samplemap must match the sample identifier — sample_id when both sample_id and sample_type are present (long-format somatic, see assets/stub/somatic_samplemap_mastersheet.csv), otherwise id (see assets/stub/samplemap_idtumi_mastersheet.csv):

id,samplemap
GM24385_unedited,assets/stub/demux_fastq/Samplemap2.umi.csv

id,sample_id,sample_type,samplemap
KDM7A-KO-CART,CD34-CART-DNA,normal,assets/stub/demux_fastq/Samplemap2.somatic.csv
KDM7A-KO-CART,KDM7A-KO-CART-DNA,tumor,assets/stub/demux_fastq/Samplemap2.somatic.csv

---

Run Command

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,ris2,dragen4,<analysis-profile> \
  --input /path/to/samplesheet.csv \
  --outdir /path/to/results/

Profiles (stack as comma-separated list)

Profile	Purpose
dhslab	dhslab group settings (queue, paths)
ris	WashU compute1, LSF scheduler
ris2	WashU compute2, SLURM scheduler
dragen4	On-premise DRAGEN 4.3.6 hardware
dhslabdragenaws	AWS Batch with DRAGEN 4.4.6
alignonly	Alignment-only workflow
germline	Germline variant calling
somaticheme	Somatic heme tumor/normal
somaticsolid	Somatic solid tumor/normal
tumoronlyheme	Tumor-only heme
rnaseq	RNA-seq
bsseq	Bisulfite/methylation
idtumi	IDT UMI adapter mode

---

Compute Profiles

WashU RIS compute1 (LSF)

-profile dhslab,ris,dragen4,<analysis-profile>

Requires --user_group, --queue, and --job_group_name (set in dhslab profile or on command line).

WashU RIS compute2 (SLURM)

-profile dhslab,ris2,dragen4,<analysis-profile>

Uses condo-dspencer partition by default. Override with --slurm_partition.

AWS Batch

nextflow secrets set AWS_ACCESS_KEY <key>
nextflow secrets set AWS_SECRET_KEY <secret>
nextflow secrets set AWS_DRAGEN_USER <dragen_user>
nextflow secrets set AWS_DRAGEN_PASSWORD <dragen_password>

nextflow run dhslab/dragenflow -r main \
  -profile dhslab,dhslabdragenaws,<analysis-profile> \
  --input samplesheet.csv \
  --outdir s3://your-bucket/results/ \
  -bucket-dir s3://dhslab-dragen-data/work/

Uses DRAGEN 4.4.6 on AWS with Wave/Fusion enabled.

---

Testing with Example Data

assets/stub/ contains small real FASTQ/CRAM/BAM files plus pre-built samplesheets ("mastersheets") that exercise every input pathway described above, so the pipeline can be exercised end-to-end without production data.

File	Input type	Pairs with
reads_wgs_mastersheet.csv	Paired FASTQs	alignonly, germline
fastq_list_wgs_mastersheet.csv	DRAGEN fastq_list CSV	alignonly, germline
bam_wgs_test_mastersheet.csv	BAM for realignment	alignonly, germline
cram_fastq_wgs_mastersheet.csv	CRAM + FASTQ merged per sample	alignonly, germline
reads_rnaseq_mastersheet.csv	Paired FASTQs	rnaseq
reads_idtumi_mastersheet.csv	Paired UMI FASTQs	idtumi (+ tumoronlyheme/somaticheme)
samplemap_idtumi_mastersheet.csv	MGI samplemap	idtumi (+ tumoronlyheme/somaticheme)
somatic_reads_mastersheet.csv	Tumor/normal FASTQs, long format	somaticheme, somaticsolid
somatic_reads_wide_mastersheet.csv	Tumor/normal FASTQs, wide format	somaticheme, somaticsolid
somatic_samplemap_mastersheet.csv	Tumor/normal MGI samplemap, long format	somaticheme, somaticsolid
somatic_samplemap_wide_mastersheet.csv	Tumor/normal MGI samplemap, wide format	somaticheme, somaticsolid

cram_wgs_test_mastersheet.csv currently points to a CRAM filename that isn't present in assets/stub/crams/ — use cram_fastq_wgs_mastersheet.csv or bam_wgs_test_mastersheet.csv for a working CRAM/BAM realignment example instead.

assets/stub/dragen_path/ holds a complete set of pre-computed DRAGEN outputs for 10929-BM-lib1 (VCFs, metrics, CRAM), useful as a reference for expected output file naming/contents.

-profile stub / -stub-run

For quick validation of pipeline logic (channel wiring, samplesheet parsing, module I/O) without DRAGEN hardware or a license, add the stub profile and Nextflow's -stub-run flag. Every process returns canned output instead of actually running:

nextflow run main.nf -profile stub,alignonly -stub-run \
  --input assets/stub/cram_fastq_wgs_mastersheet.csv \
  --outdir results_test

Example runs against real DRAGEN hardware

# Germline WGS from paired FASTQs
nextflow run main.nf -profile dhslab,ris2,dragen4,germline \
  --input assets/stub/reads_wgs_mastersheet.csv \
  --outdir results_test

# Somatic heme tumor/normal via MGI samplemap (wide format)
nextflow run main.nf -profile dhslab,ris2,dragen4,somaticheme \
  --input assets/stub/somatic_samplemap_wide_mastersheet.csv \
  --outdir results_somatic

# RNA-seq
nextflow run main.nf -profile dhslab,ris2,dragen4,rnaseq \
  --input assets/stub/reads_rnaseq_mastersheet.csv \
  --outdir results_rnatest

# IDT UMI adapter mode, tumor-only heme calling, restricted to a target BED
nextflow run main.nf -profile dhslab,ris2,dragen4,idtumi,tumoronlyheme \
  --input assets/stub/reads_idtumi_mastersheet.csv \
  --target_bedfile /path/to/targets.bed \
  --outdir results_umi_test

Add --run_dragen false to any run to execute everything up to (but not including) the DRAGEN call — useful for checking that inputs were parsed and staged correctly before committing compute time.

---

Key Parameters

Parameter	Default	Description
--input	—	Path to samplesheet CSV (required)
--outdir	—	Output directory (required)
--refdir	hg38 dragen_hg38v4.3.6	DRAGEN hash-table reference directory
--fasta	hg38_mgi_patch.fa	FASTA reference (for CRAM decoding, VEP)
--target_bedfile	null	Target BED for exome/panel; enables exome SV/CNV calling
--extra_dragen_args	null	Additional raw arguments passed to dragen
--mark_duplicates	true	Enable duplicate marking
--alignment_file_format	CRAM	Output format: CRAM or BAM
--umi	null	UMI library type (e.g., random-simplex)
--readfamilysize	3	Minimum UMI family size
--liquid_tumor	false	Enable liquid tumor UMI mode
--solid_tumor	false	Enable solid tumor UMI mode
--dux4caller	false	Enable DRAGEN DUX4 fusion caller
--use_nirvana	true	Use Nirvana for variant annotation (instead of dbSNP only)
--vepcache	VEP113_cache	VEP cache directory
--hotspot_vcf	null	Somatic hotspot VCF
--hotspot_bed	null	BED file to generate hotspot VCF on the fly
--snv_noisefile	null	Systematic noise BED for SNV filtering
--sv_noisefile	null	Systematic noise BEDPE for SV filtering
--cnv_population_vcf	null	Population BAF VCF for CNV calling
--dragen_cnv_filter_length	null	Minimum CNV length filter (bp)
--dragen_cnv_merge_distance	null	CNV merge distance (bp)
--downsample_rna	null	Downsample RNA reads (reads count)
--run_dragen	true	Set to false to skip DRAGEN (for testing)

---

Outputs

Results are written per-sample to <outdir>/<sample_id>/. DRAGEN output files are copied directly from the DRAGEN output directory.

Mode	Key outputs
All	<id>.cram / <id>.bam, alignment QC metrics, pipeline_info/
germline / somatic	<id>.hard-filtered.vcf.gz (SNVs), <id>.sv.vcf.gz, <id>.cnv.vcf.gz, VEP-annotated VCFs and TSV tables
rnaseq	<id>.quant.genes.sf, <id>.quant.sf, <id>.fusion_candidates.final, annotated expression tables
bsseq	<id>.methylation_call_file.gz, methylation BED, BigWig
All	multiqc_report.html

---

Reference Files

Default paths (hg38, dhslab group storage):

Resource	Path
DRAGEN reference (DNA)	/storage2/fs1/dspencer/Active/shared/refdata/hg38/dragen_hg38v4.3.6
DRAGEN reference (5mC)	/storage2/fs1/dspencer/Active/shared/refdata/hg38/dragen_hg38_5mCv4.3.6
FASTA	/storage2/fs1/dspencer/Active/shared/refdata/hg38/sequence/hg38_mgi_patch.fa
dbSNP	/storage2/fs1/dspencer/Active/shared/refdata/hg38/dragenfiles/dbsnp.vcf.gz
VEP cache	/storage2/fs1/dspencer/Active/shared/refdata/hg38/VEP113_cache
Nirvana annotation	/storage2/fs1/dspencer/Active/shared/refdata/hg38/dragenfiles/nirvana_annotation_data_323
SNV noise (heme WGS)	/storage2/fs1/dspencer/Active/shared/refdata/hg38/dragenfiles/IDPF_WGS_hg38_v.2.0.0_systematic_noise.snv.bed.gz
SV noise (heme WGS)	/storage2/fs1/dspencer/Active/shared/refdata/hg38/dragenfiles/WGS_FF_Heme_hg38_v3.1.0_systematic_noise.sv.bedpe.gz
Population BAF VCF	/storage2/fs1/dspencer/Active/shared/refdata/hg38/dragenfiles/1000G_phase1.snps.high_confidence.hg38.vcf.gz
Ensembl GTF (RNA)	/storage2/fs1/dspencer/Active/shared/refdata/hg38/ensemble/Homo_sapiens.GRCh38.105.chr.sorted.gtf.gz
Cytobands	assets/data/hg38.cytoBandIdeo.bed.gz