Canary is a self contained amplicon NGS pipeline that takes a pair of zipped Fastq files and generates an annotated VCF file of variants. Canary correctly describes variants in HGVS nomenclature by making a number of calls to the Mutalyzer.nl website to render them with a Refseq transcript as an HGVSg, HGVSc and HGVSp and in their most 3' form. The preferred Refseq transcript for a gene is taken from a gene to transcript mapping file which may edited by the user if required. Complex multi-nucleotide variants are correctly rendered as delins variants in their most parsimonious form to save error-prone manual interpretation of these variants - a common cause of variant description error.
Clinical diagnostics is being transformed by the technology capable of analysing patient DNA at the nucleotide level.
Amplicon sequencing is a cost effective way of deeply sequencing a panel of genes of interest. Canary has been implemented for use with a number of clinical cancer diagnostic panels at the Peter MacCallum Cancer Centre, the largest cancer only hospital in Australia.
The accuracy, turn around time and reproducibility of clinical diagnostic sequencing relies heavily on bioinformatics pipelines to convert raw sequencing data into meaningful variants that can be curated for patient reporting. Canary is a clinically tested tool which performs the key pipeline tasks of:
- alignment of amplicon reads,
- calling of variants,
- 3' shifting,
- coalescing of MNPs (multi-nucleotide polymorphisms),
- Refseq transcript selection,
- rendering of variants into HGVS nomenclature and
- annotation of the VCF file with MyVariant.info attributes.
Canary takes advantage of many open-source and public Java libraries to implement an enterprise-grade application suitable for clinical use.
It is implemented in Groovy (a Java byte compatible language) and runs on any computer with Java installed.
Canary can be installed with a single command via Docker:
% docker run -v /tmp/data:/canary.data dockercanary/canary
Docker can be downloaded and installed from here and supports many platforms such as Mac, Linux, Windows, Windows Server etc and allows Canary to be run without dependancy issues (apart from Docker). The first run will download and cache a dockerised Canary image built from the source code in this repository.
The docker container itself is built from a Linux image which can be accessed directly at the shell level with:
% docker run -it --entrypoint 'bash' dockercanary/canary
This repository contains a pre-built uber-jar bundling all dependencies for Canary. Cloning the repository will download the uber-jar ready for running the example data.
If you are more adventurous, the repository also contains all the dependencies (listed below) to build Canary. From the install directory, run:
% git clone https://github.com/PapenfussLab/Canary.git
% cd Canary
% export CANARY_HOME=/canary/target/dir # replace with where you want to install Canary - must be absolute path - starts with '/'
% mkdir -p $CANARY_HOME
% gradle uploadArchives
% PATH=$CANARY_HOME/bin:$PATH
There are a number of dependencies including the following
- Java JDK 1.7 from Oracle
- Groovy 2.1.9 from here
- Gradle for building Canary (we use 1.10 at time of writing) Gradle 1.10
- Genome Analysis Toolkit (Currently GATK 3.3) and the Sting utility JAR (currently 2.1.8) available from here
- The PathOS Core library available from PathosCore-all-1.3.jar and maintained here.
- JNI wrapper to the striped Smith-Waterman alignment library SSW see here
Note that using --normalise flag, will also require loading a reference genome (see here)
The /opt/Canary/bin directory in the docker image contains an example shell script runCanary.sh for running Canary against sample FASTQ reads files in the /opt/Canary/Fastq directory. These reads were generated on
an Illumina MiSeq platform using the TruSeq assay, a 48 gene, targeted cancer amplicon panel from Illumina. The seqencing yielded 771,606 reads for this sample of which 93.5% were mapped to the amplicons at a coverage of ~1000X.
To run Canary with demo data, output goes to /tmp/data. The first run will download the canary docker image from the docker repo
% docker run -v /tmp/data:/canary.data dockercanary/canary
To run Canary with longer demo data (~3 minutes), output goes to /tmp/data
% docker run -v /tmp/data:/canary.data -e CANARY_OPTS="LONG" dockercanary/canary
2017-03-24 01:48:52,440 [main] INFO org.petermac.pathos.pipeline.Canary - Canary [--mutalyzer, https://mutalyzer.nl, --amplicon, /opt/Canary/Amplicon/amplicon.fa, --primers, /opt/Canary/Amplicon/amplicon.primers.tsv, --transcript, /opt/Canary/etc/transcript.tsv, --columns, /opt/Canary/etc/cols, --reads, 1, --normalise, canary.norm.vcf, --tsv, outvcf.tsv, --output, canary.tsv, --vcf, canary.vcf, --bam, canary.bam, /opt/Canary/Fastq/14M6168_AACCCCTC-TAGACCTA_L001_R1_001.fastq.gz, /opt/Canary/Fastq/14M6168_AACCCCTC-TAGACCTA_L001_R2_001.fastq.gz]
2017-03-24 01:48:52,836 [main] INFO org.petermac.pathos.pipeline.Canary - Loaded 14808 gene/transcripts from /opt/Canary/etc/transcript.tsv
Using PathOS Configuration File [/opt/Canary/etc/canary.properties] PathOS Home [/opt/Canary]
2017-03-24 01:48:52,890 [main] INFO org.petermac.util.SmithWaterman - Loading SmithWaterman JNI Library from /opt/Canary/lib/libsswjni.jnilib
2017-03-24 01:48:52,945 [main] INFO org.petermac.pathos.pipeline.Canary - Found Reads file: readlen=150 filesize=30495012 estimated reads=392429
2017-03-24 01:48:52,963 [main] INFO org.petermac.pathos.pipeline.Canary - Found 221 Amplicons
2017-03-24 01:48:53,198 [main] INFO org.petermac.pathos.pipeline.Canary - PIK3CA4_11.chr3.178936074.178936095_tile_1 has fwd rev matches to Off_target_1_PIK3CA4_11.chr3.178936074.178936095_tile_1-PIK3CA4_11.chr3.178936074.178936095_tile_1
2017-03-24 01:48:53,198 [main] INFO org.petermac.pathos.pipeline.Canary - PIK3CA12.chr3.178938860.178938860_tile_1 has fwd rev matches to Off_target_2_PIK3CA12.chr3.178938860.178938860_tile_1-PIK3CA12.chr3.178938860.178938860_tile_1
2017-03-24 01:48:53,204 [main] INFO org.petermac.pathos.pipeline.Canary - PTEN13.chr10.89720716.89720852_tile_1 has fwd rev matches to Off_target_4_PTEN13.chr10.89720716.89720852_tile_1-PTEN13.chr10.89720716.89720852_tile_1
2017-03-24 01:48:53,208 [main] INFO org.petermac.pathos.pipeline.Canary - Using 212 Amplicons
2017-03-24 01:48:56,023 [main] WARN org.petermac.pathos.pipeline.Canary - Maximum estimated read limit reached: 3924 read pairs
2017-03-24 01:48:56,178 [main] INFO org.petermac.pathos.pipeline.Canary - Found 18 events
2017-03-24 01:48:56,206 [main] INFO org.petermac.pathos.pipeline.Canary - Found 18 variants
2017-03-24 01:49:20,385 [main] INFO org.petermac.pathos.pipeline.Mutalyzer - Mutalyzer: No Proxy set
2017-03-24 01:49:20,385 [main] INFO org.petermac.pathos.pipeline.Mutalyzer - Mutalyzer: No Proxy set
2017-03-24 01:49:20,416 [main] INFO org.petermac.pathos.pipeline.MutalyzerUtil - Loaded 18 variants from canary.vcf
2017-03-24 01:49:24,261 [main] INFO org.petermac.pathos.pipeline.Mutalyzer - Job running: retrieve with: https://mutalyzer.nl/batch-job-result/batch-job-880e8efe-ed82-4327-826e-cfddca18bb12.txt
2017-03-24 01:49:35,992 [main] INFO org.petermac.pathos.pipeline.Mutalyzer - Job running: retrieve with: https://mutalyzer.nl/batch-job-result/batch-job-ce7719c2-4224-40b9-9b7e-2dbfbc2fc413.txt
2017-03-24 01:49:49,739 [main] INFO org.petermac.pathos.pipeline.MutalyzerUtil - Found 0 redesignated variants
2017-03-24 01:49:49,986 [main] INFO org.petermac.pathos.pipeline.MutalyzerUtil - convertVcf(canary.vcf): In 18 Out 18
2017-03-24 01:49:50,115 [main] INFO org.petermac.pathos.pipeline.Canary - Done: processed 31400 lines 3925 read pairs into canary.vcf in 57 seconds
Typical parameters for Canary:
% Canary \
--mutalyzer 'https://mutalyzer.nl' \
--amplicon $CANARY_HOME/Amplicon/amplicon.fa \
--primers $CANARY_HOME/Amplicon/amplicon.primers.tsv \
--transcript $CANARY_HOME/etc/transcript.tsv \
--columns $CANARY_HOME/etc/cols \
--reads 10 \
--complex \
--output out.canary.tsv \
--vcf out.canary.vcf \
--bam out.canary.bam \
--normalise out.norm.vcf \
--annotation myvariant.txt \
--tsv out.norm.tsv \
$CANARY_HOME/Fastq/*R1_001.fastq.gz \
$CANARY_HOME/Fastq/*R2_001.fastq.gz
There are a number of files that make Canary work. These are referenced in the bin/Canary shell script and the properties file at etc/canary.properties.
The location of logging can be configured by editing the properties file lib/log4j.properties.
The amplicons generating the reads need to be described to Canary with two files. A fasta file (see Amplicon/amplicon.fa) containing the genomic sequences of the amplicons including primers. A second tab-delimited file describing the genomic position of the amplicons is required with the following format. (see Amplicon/amplicon.primers.tsv)
| Column | Description | Example |
|---|---|---|
| 1 | Genomic coordinates of amplicon (including primers) | 1:43814982-43815163 |
| 2 | Length of start primer (bp) | 24 |
| 3 | Length of end primer (bp) | 26 |
| 4 | Name of amplicon (Off target amplicons must start with "Off") | MPL1_2.chr1.43815008.43815009_tile_1 |
When using the --normalise flag, Canary will need access to a full reference genome FASTA file for correctly annotation 3' shifted variants. The location of this file is set by the properties file located in etc/canary.properties. Update the genome.path property to point to the genome reference. These can be obtained from 1000g for various genome builds.
Also required for the --normalise flag is a file mapping genes to refseq transcripts. A default mapping is provided here but may be edited if other transcripts are preferred. We maintain this mapping as a curated list of preferred clinical transcripts that should be reported on. It is currently cancer centric reflecting its pedigree as a clinical cancer reporting tool. Mantaining this mapping allows for consistent reporting of transcripts that should reflect the most common literature reported transcript for a gene. While this is not suitable for all circumstances, it is a necessary simplification for routine clinical practices.
When using the --tsv flag, a tab separated variable file is generated to expand out all VCF Info and Format fields as a tabular file suitable for loading into spreadsheets and databases. The fields to be rendered as a TSV file can be selected by using the --cols flag and providing a file of VCF field names to output to TSV. One per line with an optional alias after a comma if VCF field names needs to be renamed in the TSV header. (see cols).
Ken Doig, Bioinformatics, Cancer Research Department, Data Scientist, Molecular Pathology Department Peter MacCallum Cancer Centre, Victorian Comprehensive Cancer Centre Building 305 Grattan Street, Melbourne Victoria 3000 Australia Ph: +61 411 225 178 Mail: [email protected]