This workflow will download all VCF (SNV, SV, INDEL) files for all analysis pipelines (Broad, DKFZ/EMBL, Muse, Sanger) and all BAM files (normal and all tumour BAMs) to perform the various analysis tasks.

This workflow will perform three tasks:

• OxoG filtering
• Mini-bam generation
• Annotation

Once these tasks are complete, the results will be rsynced from the worker to a server, whose URL can be specified in the INI file.

This is a component by the Broad institute that will perform filtering on VCF files.

Mini-bam files are produced by a program called variantbam.

This is Jonathan Dursi's Annotator. See:

• https://hub.docker.com/r/ljdursi/pcawg-annotate/
• https://github.com/ljdursi/sga-annotate-docker

This command will just compile and package the Java portion of the workflow:

mvn clean package

To produce a full SeqWare bundle, you will need to use SeqWare:

seqware bundle package --dir target/Workflow_Bundle_OxoGWrapper_*_SeqWare_1.1.2

1. Ensure that you have a valid download credetial files in ~/.gnos. For more details, see the sections on download methods.
2. Ensure that you have a valid git pem file in ~/.gnos. See here, for more details: https://help.github.com/articles/generating-an-ssh-key/
3. Call the workflow like this:

	docker run --rm -v /datastore/:/datastore/ \
		-v /workflows/Workflow_Bundle_OxoGWrapper_3.0.2_SeqWare_1.1.2/:/workflow/ \
		-v /var/run/docker.sock:/var/run/docker.sock \
		-v /home/ubuntu/.gnos/:/home/ubuntu/.gnos/ \
		-v /home/ubuntu/SomeIniFile.INI:/ini \
		pancancer/seqware_whitestar_pancancer:1.1.2-actual-java8 \
			/bin/bash -c "seqware bundle launch --dir /workflow --ini /ini --no-metadata --engine whitestar-parallel"

This will build a new image which contains the workflow as a SeqWare bundle:

docker build -t pancancer/pcawg-oxog-wrapper-workflow:x.x.x .

(replace the x.x.x with the correct version number)

1. Ensure that you have a valid download credetial files in ~/.gnos. For more details, see the sections on download methods.
2. Ensure that you have a valid git pem file in ~/.gnos. See here, for more details: https://help.github.com/articles/generating-an-ssh-key/
3. Call the workflow like this (note: you do not have to mount the workflows directory in this case, because the workflow is already inside the container):

sudo docker run --rm -v /datastore:/datastore \
		-v /var/run/docker.sock:/var/run/docker.sock \
		-v /home/ubuntu/MyINIFile.INI:/ini \
		-v /home/ubuntu/.gnos/:/home/seqware/.gnos \
	pancancer/pcawg-oxog-wrapper-workflow:x.x.x \
		seqware bundle launch --dir /workflow --ini /ini --no-metadata --engine whitestar-parallel

The INI must reference the object IDs and filenames of INDEL, SNV, and SV VCFs from the Sanger, Broad, Muse (SNV only), and DKFZ/EMBL pipelines. Populating such an INI file is a lot of work and since most of the necessary information is already in the JSON file (produced by Linda/Junjun), you can use the INIGenerator to produce an INI for you. It works like this:

cd /workflows/Workflow_Bundle_OxoGWrapper_3.0.2_SeqWare_1.1.2/Workflow_Bundle_OxoGWrapper/2.0.6/
java -cp ./classes:./bin com.github.seqware.INIGenerator ~/BTCA-SG.BTCA_donor_A153.json

Other fields that are useful to populate in the INI:

• gitMoveTestMode - Set it to true to prevent file operations from being propagated back to github. This is really only intended for testing purposes.
• storageSource - This tells the ICGC storage client where to download from. Options are aws or collab.
• skipDownload - Set to true to skip file downloading. Useful if you can set up the environment with the files ahead of time.
• skipUpload - Set to true to skip the upload process.
• skipOxoG - Skip the OxoG filtering process. Useful if OxoG has already run on this data.
• skipVariantBam - Skip running variant to produce the minibams. Useful if you do not need to run this step.
• skipAnnotation - Skip running the annotation process. Useful if you have already run this step.
• refFile - Specify the reference fasta file to use in the workflow. Must be in a directory relative to /datastore/refdata/. Defaults to public/Homo_sapiens_assembly19.fasta.
• uploadKey - The path to the key that will be used when uploading to the rsync server.
• gnosKey - The path to the key that will be used when talking to GNOS to generate gto files and metadata.
• rsyncKey - The path to the key that will be used when the results are rsynced.
• uploadURL - The URL to use in the rsync command at the end of the workflow, such as: someUser@10.10.10.10:~/incoming/oxog_results/
• vcfDownloadMethod - Which method to use to download VCF files. Default will be to use the icgc storage client (icgcStorageClient), but you can also specify gtdownload or s3 (which will use aws s3).
• bamDownloadMethod - Which method to use to download BAM files. Default will be to use the icgc storage client (icgcStorageClient), but you can also specify gtdownload or s3 (which will use aws s3).
• smufinDownloadMethod - Which method to use to download a pipeline, such as smufin. If this is not specified, then vcfDownloadMethod will be used.
• storageSource - Where to download files from, if the downloadMethod is icgcStorageClient. Defaults to collab but if you are using the icgc storage client in AWS, you will want to specify aws.
• allowMissingFiles - Set to true if you know that some of the filesets from earlier pipelines might be missing a VCF or two (NOTE: it is assumed that you will have at least one VCF from each pipeline). Only VCFs can be missing. Missing BAM files will cause the workflow to fail.

There are four methods that this workflow can use to obtain input files:

• icgc-storage-client
• gtdownload
• aws CLI
• filesystem copy

This is the default download method. To use this download method, specify downloadMethod=icgcStorageClient in the INI file. You must have a collab.token file that is in the ~/.gnos directory of the machine that will run the workflow. To learn more about using the icgc-storage-client, see here: http://docs.icgc.org/cloud/guide/#storage-client-usage

This will use the gtdownload tool, as found in the docker container pancancer/pancancer_upload_download. To use this download method, specify downloadMethod=gtdownload in the INI file. You must have your GNOS key files available in the ~/.gnos directory. This workflow allows you to specify separate keys for BAMs and VCFs in case they are located on different servers with different credentials. You must reference the GNOS key files in the INI file like this:

gtDownloadBamKey=/home/ubuntu/.gnos/bamDownloadKey.key
gtDownloadVcfKey=/home/ubuntu/.gnos/VCFDownloadKey.key

You must also ensure that your INI file contains GNOS repos for each file set you wish to download. The INI Generator that is packaged with this workflow should do this for you, if the information is already available in its input JSON file. It should look something like this:

sanger_download_url=https://gtrepo-osdc-icgc.annailabs.com/cghub/data/analysis/download/some-long-uuid-here

This will use Amazon's AWS CLI tool to download from an S3 URL. Currently, it can only download from the OICR Collaboratory buckets in S3. To use this download method, specify downloadMethod=s3 in the INI file. You must have valid credentials to download from this location. Place your credentials file in ~/.gnos so that the workflow can have access to it when it downloads.

This will simply copy the files from a specied base path. This can be useful if you have a VM base image with input files built into it, or if files are stored on a shared network drive. You must specify fileSystemSourcePath in your INI file as the root from where files will be copied. You should use the filename as the file object ID when your download is done as a filesystem copy.

This workflow uses git to track the state of executing jobs.

To do this, a repo must be set up with a directory that contains the proper state-directories. When a JSON file is in a specific directory, in indicates the state that the workflow running that JSOn file is in. The necessary directories are:

• queued-jobs
• downloading-jobs
• running-jobs
• uploading-jobs
• completed-jobs
• failed-jobs

The JSON file for a job must first be in the queued-jobs directory. The workflow will then move it to the downloading-jobs directory. When the download phase of the workflow completes, the file is moved to the running-jobs directory. When the main processes have finished and the upload is in progress, the file will be moved to the uploading-jobs directory. When upload completes successfully, the file will be moved to the completed-jobs directory. If the workflow fails, the file may be moved to the failed-jobs directory. If you re-run a workflow on a worker manually, be aware that you'll have to move the JSON file back to the directory that the workflow expects it to be in.

The inputs to this wrapper workflow are VCFs and BAMs.

VCFs and VCF index files must be located in /datastore/vcf/${workflowName} where ${workflowName} name can be one of:

• broad
• sanger
• dkfz_embl
• muse

BAMs and BAIs must be located in /datastore/bam/${bamType}/${gnosID}/ where ${bamType} must be either tumour or normal and ${gnosID} is the GNOS ID of the BAM file.

The OxoG workspace will be located in /datastore/oxog_workspace. This directory will contain files such as logs produced by OxoG. These can be useful when debugging an OxoG failure. The OxoG results will be located in /datastore/oxog_results/tumour_${tumourGnosID}, where ${tumourGnosID} is the GNOS ID of a specific tumour. When there are multiple tumours, there will be multiple OxoG results, as OxoG must be run on each normal-tumour BAM pair.

The outputs of variantbam will be located in /datastore/variantbam_results/. There will be one mini-bam and BAI for each input BAM.

The VCF files that will be rsynced at the end of this wrapper workflow can be found in /datastore/files_for_upload/. These directories contain the pass-filtered VCFs, normalized INDEL VCFs, extracted SNV-from-INDEL VCFs, OxoG VCFs, OxoG call-stats, annotated OxoG VCFs, and the index files for all VCFs. BAM files are rsynced directly from /datastore/variantbam_results/.

Metadata files are generated for later submission to GNOS (which happens as a separate process, after the results are rsynced). These files can be found under /datastore/vcf-upload-prep/ and /datastore/bam-upload-prep/.