pdxBlacklist

Patient-derived xenograft models are extremely valuable to immunology and oncology¹. However, cross-contamination between host and graft tissue is inevitable². Accordingly, sequencing data generated from these samples likely represent a mixture of mouse and human sequences.

pdxBlacklist catalogs genomic variation artifacts that arise by aligning mouse-derived short-read sequencing data to the human genome. To achieve this, mouse strain-specific short-reads generated by the Mouse Genome Project³ are aligned to the human genome, before identifying likely artifacts (e.g. SNVs and indels). The resulting variation catalog enables annotation of existing variation datasets using tools such as Vcfanno⁴.

The pdxBlacklist pipeline is based on the bcbio-nextgen pipeline, which affords great analytic flexibility in the choice of genome builds, aligners and variant callers. Currently, the defaults are set to:

• Genome: hg19
• Aligner: BWA
• Caller: VarDict⁵

To run the pdxBlacklist pipeline, execute the following:

python pdxBlacklist.py --strain NOD_ShiLtJ

where --strain specifies the mouse strain of interest. Please select from the following strain identifiers:

129P2_OlaHsd, 129S1_SvImJ, 129S5SvEvBrd, AKR_J, A_J, BALB_cJ, BTBR_T__Itpr3tf_J, BUB_BnJ, C3H_HeH, C3H_HeJ, C57BL_10J, C57BL_6NJ, C57BR_cdJ, C57L_J, C58_J, CAST_EiJ, CBA_J, DBA_1J, DBA_2J, FVB_NJ, I_LnJ, KK_HiJ, LEWES_EiJ, LP_J, MOLF_EiJ, NOD_ShiLtJ, NZB_B1NJ, NZO_HlLtJ, NZW_LacJ, PWK_PhJ, RF_J, SEA_GnJ, SPRET_EiJ, ST_bJ, WSB_EiJ, ZALENDE_EiJ, JF1_MsJ, LG_J, SJL_J, SM_J

Additional options include:

• --help show a help message and exit
• --strain Mouse Genome Project strain (see ftp://ftp-mouse.sanger.ac.uk/REL-1604-BAM/) [string]
• --cores Number of cores to use for bcbio [integer]
• --debug Debugging mode [boolean]
• --config Full path to filename of an optional BCBIO config file: for details, see http://bcbio-nextgen.readthedocs.io/en/latest/contents/configuration.html

The automatically generated default BCBIO config file is:

details:
- algorithm:
    aligner: bwa
    save_diskspace: true
    mark_duplicates: true
    realign: false
    recalibrate: false
    remove_lcr: false
    platform: illumina
    quality_format: standard
    variantcaller: vardict
    indelcaller: false
    phasing: false
  analysis: variant2
  description: NOD_ShiLtJ
  files:
  - /home/maxsalm/pdxBlacklist_output/NOD_ShiLtJ.1.fq
  - /home/maxsalm/pdxBlacklist_output/NOD_ShiLtJ.2.fq
  genome_build: hg19
fc_date: 2017/5/17
fc_name: NOD_ShiLtJ
resources:
  tmp:
    dir: /tmp
upload:
  dir: ../final

The pipeline output is written to disk in VCF format to a directory labelled by strain within the pdxBlacklist_output directory in your home directory (see BCBIO for more details on output directory structure).

Features

• A choice of mouse strains (currently 40)
• Diverse variant analysis pipelines

Recommended Minimum System Requirements

• Operating Systems: Linux
• Storage: 1Tb
• CPU: 8 cores
• Memory: 30Gb
• Software: Python (2.7), git

Installation

The install the pdxBlacklist pipeline, primarily relies on Ruffus and bcbio-nextgen. To simplify package management, the bioconda package manager is recommended. After installing conda and setting up the bioconda channel, please execute the following at the commandline:

## Setup bioconda channel
conda config --add channels conda-forge
conda config --add channels defaults
conda config --add channels r
conda config --add channels bioconda

## BCBIO Variables
# Please ammend these as required
ROOT="/usr/local/share"
GENOME="hg19"
ALIGNER="bwa"

## Install bcbio 
wget https://raw.github.com/chapmanb/bcbio-nextgen/master/scripts/bcbio_nextgen_install.py
python bcbio_nextgen_install.py $ROOT/bcbio --genomes $GENOME --aligners $ALIGNER --distribution ubuntu --tooldir=$ROOT/bcbio
bcbio_nextgen.py upgrade -u stable 

### Install Ruffus
conda install ruffus=2.6.3

### Install SRA Toolkit
conda install -c r r-xml=3.98_1.5 
conda install bioconductor-sradb=1.32.0
conda install sra-tools=2.8.1

Runtime

The scripts in this project have been developed and tested on an 8-core server with 30GB RAM running Ubuntu 14.04.5 LTS (GNU/Linux 3.19.0-74-generic x86_64). A full run of the SM_J strain completed in approximately two days.

Pre-processed blacklist

A blacklist was created by processing whole genome sequencing data generated for the Mouse Genomes Project for 18 mouse strains³. Alignment to the hg19 genome build was performed by BWA-MEM, and variants were called using VarDict: this generated a false posi-tive dataset comprising 11,119,424 SNVs, 13,77,355 indels and 2,305,881 complex variants. The merged VCF file (merged_blacklist.vcf.gz) can be downloaded from:

https://download.genego.com/data/merged_blacklist.vcf.gz https://download.genego.com/data/merged_blacklist.vcf.gz.tbi

Contribute

• Issue Tracker: https://github.com/MaxSalm/pdxBlacklist/issues
• Source Code: https://github.com/MaxSalm/pdxBlacklist

Support

If you are having issues, please let us know via the Issue Tracker and we will endeavour to resolve them as soon as possible.

Known Bugs

1. Mouse Genome Project BAMs are retrieved via FTP using the ftplib python package. At the end of file transfer of large files, ftplib occasionally hangs. If the BAM download is complete, then this problem can be resolved simply by stopping the the pdxBlacklist pipeline and re-starting it.

Planned improvements

1. An option to use a pre-specified BAM on disk
2. An option to process an Sequence Read Archive (SRA) submission directly

License

The project is licensed under the GPL license.

Citation

pdxBlacklist: Identifying artefactual variants in patient-derived xenograft samples Max Salm, Sven-Eric Schelhorn, Lee Lancashire, Thomas Grombacher bioRxiv 180752; doi: https://doi.org/10.1101/180752 https://www.biorxiv.org/content/early/2017/08/25/180752

References

1. Townsend, E. C. et al. The public repository of xenografts enables discovery and randomized phase iI-like trials in mice. Cancer cell 29, 574–586 (2016).

2. Conway, T. et al. Xenome–a tool for classifying reads from xenograft samples. Bioinformatics (Oxford, England) 28, i172–i178 (2012).

3. Keane, T. M. et al. Mouse genomic variation and its effect on phenotypes and gene regulation. Nature 477, 289–294 (2011).

4. Pedersen, B. S., Layer, R. M. & Quinlan, A. R. Vcfanno: Fast, flexible annotation of genetic variants. Genome biology 17, 118 (2016).

5. Lai, Z. et al. VarDict: A novel and versatile variant caller for next-generation sequencing in cancer research. Nucleic acids research 44, e108 (2016).