call-gSV

• Overview
• How To Run
• Flow Diagram
• Pipeline Steps
  • Discovery
    • 1. Calling Structural Variants
    • 2. Calling Copy Number Variants
    • 3. Check Output Quality
  • Regenotyping
    • 1. Regenotyping Structural Variants
    • 2. Regenotyping Copy Number Variants
• Inputs
• Outputs
• Testing and Validation
  • Test Data Set
  • Performance Validation
  • Quality Check Result Comparison
  • Human Genome Benchmarks
  • Validation Tool
• References
• License

Overview

The call-gSV nextflow pipeline, calls structural variants (SVs) and copy number variants (CNVs) utilizing Delly and Manta. Additionally, the pipeline can also regenotype previously identified SVs or CNVs with Delly. It is suitable for detecting copy-number variable deletion and tandem duplication events as well as balanced rearrangements such as inversions or reciprocal translocations and validates the output quality with BCFtools. The pipeline has been engineered to run in a 4 layer stack in a cloud-based scalable environment of CycleCloud, Slurm, Nextflow and Docker. Additionally it has been validated with the SMC-HET dataset and the GRCh38 reference genome, using paired-end FASTQ's that were back-extracted from BAMs created by BAM Surgeon.

Node Specific Config File Settings

Config File	Available Node cpus / memory	Designated Process 1; cpus / memory	Designated Process 2; cpus / memory	Designated Process 3; cpus / memory
F2.config	2 / 3 GB	call_gSV_Delly; 1 / 2 GB	call_gCNV_Delly; 1 / 2 GB	call_gSV_Manta; 1 / 2 GB*
F16.config	16 / 26 GB	call_gSV_Delly; 1 / 8 GB	call_gCNV_Delly; 1 / 8 GB	call_gSV_Manta; 6 / 8 GB
F32.config	32 / 62.8 GB	call_gSV_Delly; 1 / 15 GB	call_gCNV_Delly; 1 / 15 GB	call_gSV_Manta; 12 / 15 GB
F72.config	72 / 136.8 GB	call_gSV_Delly; 1 / 30 GB	call_gCNV_Delly; 1 / 30 GB	call_gSV_Manta; 24 / 30 GB
M64.config	64 / 950 GB	call_gSV_Delly; 1 / 60 GB	call_gCNV_Delly; 1 / 60 GB	call_gSV_Manta; 30 / 60 GB

---

* Manta SV calling wouldn't work on an F2 node due to incompatible resources. In order to test the pipeline for tasks not relevant to Manta, please set run_manta = false in the sample specific config file.

How To Run

Requirements

Currently supported Nextflow versions: v23.04.2

Run steps

Below is a summary of how to run the pipeline. See here for full instructions.

Pipelines should be run WITH A SINGLE SAMPLE AT TIME. Otherwise resource allocation and Nextflow errors could cause the pipeline to fail.

1. The recommended way of running the pipeline is to directly use the source code located here: /hot/software/pipeline/pipeline-call-gSV/Nextflow/release/, rather than cloning a copy of the pipeline.

   • The source code should never be modified when running our pipelines

2. Create a config file for input, output, and parameters. An example for a config file can be found here. See Nextflow Config File Parameters for the detailed description of each variable in the config file.

   • Do not directly modify the source template.config, but rather you should copy it from the pipeline release folder to your project-specific folder and modify it there

3. Create the input YAML using the template. See Input YAML for a detailed description.

   • Again, do not directly modify the source template YAML file. Instead, copy it from the pipeline release folder to your project-specific folder and modify it there.

4. The pipeline can be executed locally using the command below:

nextflow run path/to/main.nf -config path/to/sample-specific.config

• For example, path/to/main.nf could be: /hot/software/pipeline/pipeline-call-gSV/Nextflow/release/5.0.0-rc.1/main.nf
• path/to/sample-specific.config is the path to where you saved your project-specific copy of template.config
• path/to/input.yaml is the path to where you saved your sample-specific copy of call-gSV-input.yaml

To submit to UCLAHS-CDS's Azure cloud, use the submission script here with the command below:

python path/to/submit_nextflow_pipeline.py \
    --nextflow_script path/to/main.nf \
    --nextflow_config path/to/sample-specific.config \
    --nextflow_yaml path/to/input.yaml \
    --pipeline_run_name <sample_name> \
    --partition_type F16 \
    --email <your UCLA email, [email protected]>

In the above command, the partition type can be changed based on the size of the dataset. An F16 node is generally recommended for larger datasets like A-full.

Note: Because this pipeline uses an image stored in the GitHub Container Registry, you must follow the steps listed in the Docker Introduction on Confluence to set up a PAT for your GitHub account and log into the registry on the cluster before running this pipeline.

---

Flow Diagram

---

Pipeline Steps

Discovery

The "discovery" branch of the call-gSV pipeline allows you to identify germline SVs and CNVs utilizing either Delly or Manta. After variants are identified, basic quality checks are performed on the outputs of the processes.

1. Calling Structural Variants

The first step of the pipeline requires an aligned and sorted BAM file and BAM index as an input for variant calling with Delly or Manta. Delly combines short-range and long-range paired-end mapping and split-read analysis for the discovery of balanced and unbalanced SVs at single-nucleotide breakpoint resolution (deletions, tandem duplications, inversions and translocations.) SVs are called, annotated and merged into a single BCF file. A default exclude map of Delly can be incorporated as an input which removes the telomeric and centromeric regions of all human chromosomes since these regions cannot be accurately analyzed with short-read data. Manta calls SVs and indels from mapped paired-end sequencing reads. It is optimized for analysis of germline variation in small sets of individuals and somatic variation in tumor/normal sample pairs. Manta discovers, assembles and scores large-scale SVs, medium-sized indels and large insertions within a single efficient workflow.

Currently the following filters are applied by Delly when calling SVs. Parameters with a "call-gSV default" can be updated in the nextflow.config file.

Parameter	Delly default	call-gSV default	Description
svtype	ALL		SV type to compute (DEL, INS, DUP, INV, BND, ALL)
map-qual	1	20	Minimum paired-end (PE) mapping quality
qual-tra	20		Minimum PE quality for translocation
mad-cutoff	9		Insert size cutoff, median+s*MAD (deletions only)
minclip	25		Minimum clipping length
min-clique-size	2		Minimum PE/SR clique size
minrefsep	25		Minimum reference separation
maxreadsep	40		Maximum read separation

2. Calling Copy Number Variants

The second step of the pipeline identifies CNVs. To do this, Delly requires an aligned and sorted BAM file, as well as the BCF output from the SV calling step (to refine breakpoints) and a mappability map. Any CNVs identified are annotated and output as a single BCF file.

Currently the following filters are applied by Delly when calling CNVs. Parameters with a "call-gSV default" can be updated in the sample specific nextflow config file.

Parameter	Delly default	call-gSV default	Description
quality	10		Minimum mapping quality
ploidy	2		Baseline ploidy
sdrd	2		Minimum SD read-depth shift
cn-offset	0.100000001		Minimum CN offset
cnv-size	1000		Minimum CNV size
window-size	10000		Window size
window-offset	10000		Window offset
fraction-window	0.25		Minimum callable window fraction [0,1]
scan-window	10000		Scanning window size
fraction-unique	0.800000012		Uniqueness filter for scan windows [0,1]
mad-cutoff	3		Median + 3 * mad count cutoff
percentile	0.000500000024		Excl. extreme GC fraction

3. Check Output Quality

For Delly, VCF files are generated from the BCFs to run the vcf-validate command from VCFTools and vcfstats from RTGTools. Outputs from both provide preliminary summary statistics that can be viewed and evaluated in preparation for downstream cohort-wide re-calling and re-genotyping. In the Manta branch of the pipeline, a stats directory is generated under the specific output directory /Manta-/results/stats where information can be found regarding the SVs identified.

Regenotyping

The "regenotyping" branch of the call-gSV pipeline allows you to regenotype previously identified SVs or CNVs using Delly.

1. Regenotyping Structural Variants

Similar to the "discovery" process, the first step of the regenotyping pipeline requires an aligned and sorted BAM file, BAM index, and a merged sites BCF (from the merge-SVsites pipeline) as inputs for SV regenotyping with Delly. The provided sample is genotyped with the merged sites list. SVs are annotated and merged into a single BCF file. A default exclude map of Delly can be incorporated as an input which removes the telomeric and centromeric regions of all human chromosomes since these regions cannot be accurately analyzed with short-read data.

2. Regenotyping Copy Number Variants

The second possible step of the regenotyping pipeline requires an aligned and sorted BAM file, BAM index, and a merged sites BCF as an input, as well as the BCF output from the initial SV calling (to refine breakpoints) and a mappability map. Any CNVs identified are annotated and output as a single BCF file.

---

Inputs

Input YAML

Field	Type	Description
sample_id	string	Sample ID
normal	path	Set to absolute path to input BAM

---
input:
  BAM:
    normal:
      - "/path/to/input/BAM"

Note: This pipeline is designed to detect germline SVs. To maintain consistency with other Boutros Lab Nextflow pipelines, the input YAML format mirrors that of other somatic or germline variant calling pipelines. However, it's important to note that the sample type tags, whether labeled as normal or tumor, do NOT influence the germline SV/CNV calling processes in this pipeline.

Nextflow Config File Parameters

Input Parameter	Required	Type	Description
dataset_id	yes	string	Boutros lab dataset id.
blcds_registered_dataset	yes	boolean	Affirms if dataset should be registered in the Boutros Lab Data registry. Default value is false.
variant_type	yes	list	List containing variant types to call. Default is ["gSV", "gCNV"]
run_discovery	yes	boolean	Specifies whether or not to run the "disovery" branch of the pipeline. Default value is true. (either run_discovery or run_regenotyping must be true)
run_regenotyping	yes	boolean	Specifies whether or not to run the "regenotyping" branch of the pipeline. Default value is false. (either run_discovery or run_regenotyping must be true)
merged_sites	yes	path	The path to the merged sites.bcf file. Must be populated if running the regenotyping branch.
run_delly	true	boolean	Whether or not the workflow should run Delly (either run_delly or run_manta must be set to true)
run_manta	true	boolean	Whether or not the workflow should run Manta (either run_delly or run_manta must be set to true)
run_qc	no	boolean	Optional parameter to indicate whether subsequent quality checks should be run on Delly outputs. Default value is false.
reference_fasta	yes	path	Absolute path to the reference genome FASTA file. The reference genome is used by Delly for SV calling.
exclusion_file	yes	path	Absolute path to the delly reference genome exclusion file utilized to remove suggested regions for SV calling. On Slurm, an HG38 exclusion file is located at /hot/resource/tool-specific-input/Delly/hg38/human.hg38.excl.tsv
mappability_map	yes	path	Absolute path to the delly mappability map to support GC and mappability fragment correction in CNV calling
map_qual	no	path	minimum paired-end (PE) mapping quaility threshold for Delly.
save_intermediate_files	yes	boolean	Optional parameter to indicate whether intermediate files will be saved. Default value is false.
output_dir	yes	path	Absolute path to the directory where the output files to be saved.
work_dir	optional	path	Path of working directory for Nextflow. When included in the sample config file, Nextflow intermediate files and logs will be saved to this directory. With ucla_cds, the default is /scratch and should only be changed for testing/development. Changing this directory to /hot or /tmp can lead to high server latency and potential disk space limitations, respectively.
docker_container_registry	optional	string	Registry containing tool Docker images. Default: ghcr.io/uclahs-cds

An example of the NextFlow Input Parameters Config file can be found here.

Base resource allocation updaters

To optionally update the base resource (cpus or memory) allocations for processes, use the following structure and add the necessary parts to the input.config file. The default allocations can be found in the node-specific config files

base_resource_update {
    memory = [
        [['process_name', 'process_name2'], <multiplier for resource>],
        [['process_name3', 'process_name4'], <different multiplier for resource>]
    ]
    cpus = [
        [['process_name', 'process_name2'], <multiplier for resource>],
        [['process_name3', 'process_name4'], <different multiplier for resource>]
    ]
}

Note Resource updates will be applied in the order they're provided so if a process is included twice in the memory list, it will be updated twice in the order it's given. Examples:

• To double memory of all processes:

base_resource_update {
    memory = [
        [[], 2]
    ]
}

• To double memory for call_gSV_Delly and triple memory for run_validate_PipeVal and call_gSV_Manta:

base_resource_update {
    memory = [
        ['call_gSV_Delly', 2],
        [['run_validate_PipeVal', 'call_gSV_Manta'], 3]
    ]
}

• To double CPUs and memory for call_gSV_Manta and double memory for run_validate_PipeVal:

base_resource_update {
    cpus = [
        ['call_gSV_Manta', 2]
    ]
    memory = [
        [['call_gSV_Manta', 'run_validate_PipeVal'], 2]
    ]
}

---

Outputs

Output	Description
.bcf	Binary VCF output format with SVs if found.
.vcf	VCF output format with SVs if found. If output by Manta, these VCFs will be compressed.
.bcf.csi	CSI-format index for BAM files.
.validate.txt	output file from vcf-validator.
.stats.txt	output file from RTG Tools.
report.html, timeline.html and trace.txt	A Nextflow report, timeline and trace files.
*.log.command.*	Process and sample specific logging files created by nextflow.
*.sha512	generates SHA-512 hash to validate file integrity.

---

Testing and Validation

Test Data Set

Testing was performed leveraging aligned and sorted BAMs generated using bwa-mem2-2.1 against reference GRCh38 (SMC-HET was aligned against hs37d5):

• A-mini: BWA-MEM2-2.1_TEST0000000_TWGSAMIN000001-T001-S01-F.bam
• A-partial: BWA-MEM2-2.1_TEST0000000_TWGSAPRT000001-T001-S01-F.bam
• A-full: a-full-CPCG0196-B1.bam*
• A-partial: CPCG0196-B1-downsampled-a-partial-sorted.bam*
• SMC-HET: HG002.N.bam

* In Delly v1.1.3, a coverage check has been introduced which checks for coverage quality in a given window before CNV calling. Successful CNV calling was observed on samples with coverages across the genome, such as, a-full-CPCG0196-B1.bam and CPCG0196-B1-downsampled-a-partial-sorted.bam (WGS samples). For more details, please refer to Discussion #64.

Test runs for the A-mini/partial/full samples were performed using the following reference files

• reference_fasta: /hot/resource/reference-genome/GRCh38-BI-20160721/Homo_sapiens_assembly38.fasta
• exclusion_file: /hot/resource/tool-specific-input/Delly/GRCh38/human.hg38.excl.tsv
• mappability_map: /hot/resource/tool-specific-input/Delly/GRCh38/Homo_sapiens.GRCh38.dna.primary_assembly.fa.r101.s501.blacklist.gz

Performance Validation

with Delly <= v0.9.1 in the pipeline

Testing was performed primarily in the Boutros Lab Slurm Development cluster but additional functional tests were performed on the SGE cluster on 2/26/2021 and the Slurm Covid cluster. Metrics below will be updated where relevant with additional testing and tuning outputs.

Test Case	Test Date	Node Type	Duration	CPU Hours	Virtual Memory Usage (RAM) -peak rss
A-mini	2021-02-12	F2	1m 29s	a few seconds	208.8 MB
A-partial	2021-02-10	F72	42m 5s	48.8	8.9 GB
A-full	2021-02-10	F72	7h 10m 43s	509.0	10.9 GB
SMC-HET	2021-02-12	F72	3h 9m 60s	223.5	8.9 GB

with Delly >= v1.0.3 in the pipeline

Metrics below are based on the integration of Delly v1.13 in the call-gSV pipeline.

Test Case	Test Date	Node Type	Duration	CPU Hours	Virtual Memory Usage (RAM) -peak rss
CPCG0196-B1 A-partial	2022-08-08	F72	1h 9m 15s	2.2	10.85 GB
CPCG0196-B1 A-full	2022-08-06	F72	21h 3m 19s	37.3	24.68 GB

Quality Check Result Comparison

Metric	A-mini	A-partial	A-full	SMC-HET	Source
Count Pass	3	2593	62704	15196	grep -c -w "PASS" filename.vcf -1
Count Deletion	2	1475	49433	9317	grep -c -w "SVTYPE=DEL" filename.vcf
Count Duplication	1	170	2311	1705	grep -c -w "SVTYPE=DUP" filename.vcf
Count Inversion	0	317	2801	2197	grep -c -w "SVTYPE=INV" filename.vcf
Count Translocation	0	384	7439	0	grep -c -w "SVTYPE=BND" filename.vcf
Count Insertion	0	267	1265	2059	grep -c -w "SVTYPE=INS" filename.vcf
PRECISE Calls	3	1850	11541	8267	grep -c -w "PRECISE" filename.vcf
IMPRECISE Calls	2	764	51709	7012	grep -c -w "IMPRECISE" filename.vcf
Failed Filters	0	653	44991	2619	.stats.txt
Passed Filters	3	1959	18257	12658	.stats.txt
SV breakends	0	219	1124	0	.stats.txt
Symbolic SVs	2	1559	12500	11156	.stats.txt
Same as reference	1	263	4595	1471	.stats.txt
Missing Genotype	0	8	38	31	.stats.txt
Total Het/Hom ratio	(2/0)	1.00 (843/845)	2.37 (9580/4044)	1.86 (7251/3905)	.stats.txt
Breakend Het/Hom ratio	(0/0)	0.84 (59/70)	13.41 (1046/78)	(0/0)	.stats.txt
Symbolic SV Het/Hom ratio	(2/0)	1.01 (784/775)	2.15 (8534/3966)	1.86 (7251/3905)	.stats.txt
Duplicate entries	0 errors total	1 error chr8:3893339	1 error chr1:16050024	1 error chr1:187464829	.validate.txt

Human Genome Benchmarks

Note, per Nature the following benchmarks exist for the human genome: “Structural variants affect more bases: the typical genome contains an estimated 2,100 to 2,500 structural variants (∼1,000 large deletions, ∼160 copy-number variants, ∼915 Alu insertions, ∼128 L1 insertions, ∼51 SVA insertions, ∼4 NUMTs, and ∼10 inversions), affecting ∼20 million bases of sequence.”

Validation Tool

Included is a template for validating your input files. For more information on the tool check out: https://github.com/uclahs-cds/public-tool-PipeVal

---

References

1. Rausch T, Zichner T, Schlattl A, Stütz AM, Benes V, Korbel JO. DELLY: structural variant discovery by integrated paired-end and split-read analysis. Bioinformatics. 2012;28(18):i333-i339. doi:10.1093/bioinformatics/bts378
2. Chen, X. et al. (2016) Manta: rapid detection of structural variants and indels for germline and cancer sequencing applications. Bioinformatics, 32, 1220-1222. doi:10.1093/bioinformatics/btv710
3. VCFtools - vcf-validator
4. Real Time Genomics RTG Tools Operations Manual - vcfstats
5. Boutros Lab -CallSV Quality Control pipeline
6. The 1000 Genomes Project Consortium., Corresponding authors., Auton, A. et al. A global reference for human genetic variation. Nature 526, 68–74 (2015). https://doi.org/10.1038/nature15393

Contributors

Please see list of Contributors at GitHub.

License

The pipeline-call-gSV is licensed under the GNU General Public License version 2. See the file LICENSE for the terms of the GNU GPL license.

The pipeline-call-gSV takes BAM and BCF files and utilizes Delly to call/regenotype gSV/gCNV.

Copyright (C) 2021-2024 University of California Los Angeles ("Boutros Lab") All rights reserved.

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.