OSG-GEM is a Pegasus workflow that utilizes Open Science Grid (OSG) resources to produce a Gene Expression Matrix (GEM) from DNA sequence files in FASTQ format. The workflow is also configured to run on Jetstream
William L. Poehlman, Mats Rynge, Chris Branton, D. Balamurugan and Frank A. Feltus. OSG-GEM: Gene Expression Matrix Construction Using the Open Science Grid. Bioinformatics and Biology Insights 2016:10 133–141 doi: 10.4137/BBI.S38193.
This workflow processes paired or single end FASTQ files to produce a matrix of normalized RNA molecule counts (FPKM). OSG-GEM also supports direct input downloads from NCBI SRA for processing. An indexed reference genome along with gene model annotation files must be obtained prior to configuring and running the workflow. The following tasks are directed by the Pegasus workflow manager:
- Splitting input FASTQ files into files containing 20,000 sequences each.
- Trimming raw sequences with Trimmomatic
- Aligning reads to the reference genome using Hisat2 or Tophat2
- Merging BAM alignment files into a single sorted BAM file using Samtools
- Quantifying RNA transcript levels using StringTie or Cufflinks
It is suggested that the user become familiar with the documentation associated with the following software packages:
The OSG-GEM workflow is designed to execute on the Open Science Grid via the OSG Connect infrastructure. Access to the system can be requested on the sign up page.
Once you have an account and have joined a project, login to the login02 submit node. This node can be accessed by ssh:
A workflow specific ssh key has to be created. This key is used for some of the data staging steps of the workflow.
$ mkdir -p ~/.ssh
$ ssh-keygen -t rsa -b 2048 -f ~/.ssh/workflow
(just hit enter when asked for a passphrase)
One you have the ~/.ssh/workflow.pub file, add it to your profile as described in https://support.opensciencegrid.org/support/solutions/articles/12000027675-generate-ssh-key-pair-and-add-the-public-key-to-your-account#step-2-add-the-public-ssh-key-to-login-node
The worklow cloned from github contains an example config file as well as example input files from the 21st chromosome of Gencode Release 24 of the GRCh38 build of the human reference genome. Two small FASTQ files containing 200,000 sequences from NCBI dataset SRR1825962 lie within the Test_data directory of the workflow. To run the test workflow, the user must copy the osg-gem.config.template file:
$ cp osg-gem.conf.example osg-gem.conf
The workflow, configured to run Hisat2 and Stringtie, can then be launched by running:
$ ./submit
From here, the user may follow our documentation to modify the software options as well as point to their own input datasets. Note that there are no test reference genome indices available for STAR, because they are too large to upload to github.
The user must provide indexed reference genome files as well as gene model annotation information prior to submitting the workflow. The user must select a reference prefix($REF_PREFIX) that will be recognized by Pegasus as well as by Hisat2 or Tophat2. In addition, information about splice sites or a reference transcriptome must be provided in order to guide accurate mapping of split input files. Once the user has downloaded a reference genome fasta file and gene annotation in GTF/GFF3 format, the following commands can be used to produce the necessary input files, using GRCh38 as an example $REF_PREFIX for Gencode Release 24 of the human reference genome:
$ hisat2-build -f GRCh38.fa GRCh38
Generate Tab delimited list of splice sites using gene model GTF file as input (Python DefaultDictionary Module necessary)
$ python hisat2_extract_splice_sites.py GRCh38-gencode.v24.annotation.gtf > GRCh38.Splice_Sites.txt
$ bowtie2-build GRCh38.fa GRCh38
$ tophat2 -G GRCh38.gencode.v24.annotation.gff3 --transcriptome-index=transcriptome_data/GRCh38 GRCh38
$ tar czf GRCh38.transcriptome_data.tar.gz transcriptome_data/
# cd into the "star_index" directory within "reference". Place all genome files here
$ cd reference/star_index
$ STAR-2.5.2b/bin/Linux_x86_64_static/STAR --runMode genomeGenerate --runThreadN 4 --genomeDir ./ --genomeFastaFiles ./GRCh38.fa --sjdbGTFfile ./GRCh38.gencode.v24.annotation.gff3
Generate Tab delimited list of splice sites using gene model GTF file as input (Python DefaultDictionary Module necessary)
$ python hisat2_extract_splice_sites.py GRCh38-gencode.v24.annotation.gtf > GRCh38.Splice_Sites.txt
Once the user has obtained necessary input files, the osg-gem.config file must be appropriately modified and reference files must be placed into the reference directory with appropriate filenames.
If the user would like to use Hisat2, the following files must be present in the reference directory:
$REF_PREFIX.fa
$REF_PREFIX.1.ht2 … $REF_PREFIX.N.ht2
$REF_PREFIX.Splice_Sites.txt
$REF_PREFIX.gff3
If the user would like to use Tophat2, the following files must be present in the reference directory:
$REF_PREFIX.fa
$REF_PREFIX.1.bt2 … $REF_PREFIX.N.bt2
$REF_PREFIX.rev.1.bt2
$REF_PREFIX.rev.2.bt2
$REF_PREFIX.transcriptome_data.tar.gz
$REF_PREFIX.gff3
If the user would like to use, STAR, the following files must be present in the reference/star_index directory:
chrLength.txt chrNameLength.txt chrName.txt chrStart.txt Genome genomeParameters.txt Log.out SA SAindex *Splice_Sites.txt
- NOTE: when using STAR, place the gff3 file in the base reference directory
OSG-GEM supports the processing of multiple input datasets into a single Gene Expression Matrix(GEM). The user may point to paired or single end FASTQ files on an OSG filesystem, or simply specify NCBI Sequence Read Archive (SRA) ID's that they would like to process. A blend of FASTQ files on OSG, as well as SRA ID's may be provided. Please note, however, that a mixture of single end and paired end reads can not be used. The user must select either Paired end or Single end reads
Each line in the config file can either be a pair of forward and reverse files, separated by a space:
input1 = forward.fastq.gz reverse.fastq.gz
Or a single fastq file (for single end reads):
input1 = test.fastq.gz
Or a single SRA ID:
input2 = DRR046893
[reference]
reference_prefix = $REF_PREFIX
[inputs]
input1 = /path_to_forward_data/TEST_1.fastq.gz ./path_to_reverse_data/TEST_2.fastq.gz or SRAID or ./path_to_fastq/TEST.fastq.gz
input2 = /path_to_forward_data/TEST2_1.fastq.gz ./path_to_reverse_data/TEST2_2.fastq.gz or SRAID or ./patch_to_fastq/TEST2.fastq.gz
[config]
single = 'True' or 'False'
paired = 'True' or 'False'
tophat2 = 'True' or 'False'
hisat2 = 'True' or 'False'
cufflinks = 'True' or 'False'
stringtie = 'True' or 'False'
If a user cloned OSG-GEM into '/stash2/user/username/GEM_test', and placed input paired-end FASTQ files for dataset 'TEST' in '/stash2/user/username/Data'. To process this dataset, along with dataset DRR046893 from NCBI SRA, using Hisat2 and StringTie with the GRCh38 build of the human reference genome, the osg-gem.config file would be modified as follows:
[reference]
reference_prefix = GRCh38
[inputs]
input1 = ./Test_data/TEST_1.fastq.gz ./Test_data/TEST_2.fastq.gz
input2 = DRR046893
[config]
single = False
paired = True
tophat2 = False
hisat2 = True
cufflinks = False
stringtie = True
Pegasus provides a set of commands to monitor workflow progress. The path to the worklow files as well as commands to monitor the workflow will print to screen upon submitting the workflow. For example:
2016.05.26 23:31:03.859 CDT: Your workflow has been started and is running in
2016.05.26 23:31:03.869 CDT: /stash2/user/username/workflows/osg-gem-x/workflow/osg-gem-x
2016.05.26 23:31:03.880 CDT: *** To monitor the workflow you can run ***
2016.05.26 23:31:03.891 CDT: pegasus-status -l /stash2/user/username/workflows/osg-gem-x/workflow/osg-gem-x
2016.05.26 23:31:03.901 CDT: *** To remove your workflow run ***
2016.05.26 23:31:03.912 CDT: pegasus-remove /stash2/user/username/workflows/osg-gem-x/workflow/osg-gem-x
Output will be transferred at the base of this directory upon completion. For example:
$ cd /stash2/user/username/workflows/osg-gem-x
$ ls
data outputs scratch workflow
$ cd outputs
$ head -1 TEST-merged_counts.fpkm
TranscriptID FPKM
To customize OSG-GEM parameters, basic understanding of the directory structure of the workflow is necessary.
This contains small FASTQ files for testing. The user may place their own data in this directory or elsehwere on the OSG filesystems.
Contains all reference genome and annotation files, as described previously.
This directory contains job wrappers for each step of the workflow. It is suggested that the user becomes familiar with the parameters set for each software to determine if they would like to make changes. If the user would like to change software parameters, they may modify the commands in the files here. Note that any changes to input filenames in the commands must match the files that are catalogued in the task-files directory (explained below)
This directory contains subdirectories for each job that utilizes specific files(eg., python script to parse StringTie output, fasta_adapters.txt file for trimmomatic).
Any files placed in these directories will be transferred to OSG compute nodes for the corresponding jobs. For example, if the user would like to use a different fasta adapters file 'NewAdapters.txt' for read trimming for the hisat2 job, they would copy this file to the hisat2 directory. Note that the job wrapper in the tools directory must now be modified to match this filename.
Contains files that may be useful to users of this workflow. Currently holds the hisat2_extract_splice_sites.py script that comes with the Hisat2 software package. This script can be used to generate a tab delimited list of splice sites from a GTF gene model file.
The base directory of the worfklow contains the osg-gem.config file, the submit script, and a pegasus configuration file.
The execution environment is catalogued in the submit script, allowing the user to alter the resources requested by the workflow.
For example:
<site handle="condorpool" arch="x86_64" os="LINUX">
<profile namespace="pegasus" key="style" >condor</profile>
<profile namespace="condor" key="universe" >vanilla</profile>
<profile namespace="condor" key="requirements" >OSGVO_OS_STRING == "RHEL 6" && HAS_MODULES == True && HAS_SCP == True && GLIDEIN_ResourceName != "Hyak" && TARGET.GLIDEIN_ResourceName =!= MY.MachineAttrGLIDEIN_ResourceName1 && TARGET.GLIDEIN_ResourceName =!= MY.MachineAttrGLIDEIN_ResourceName2 && TARGET.GLIDEIN_ResourceName =!= MY.MachineAttrGLIDEIN_ResourceName3 && TARGET.GLIDEIN_ResourceName =!= MY.MachineAttrGLIDEIN_ResourceName4</profile>
<profile namespace="condor" key="request_memory" >5 GB</profile>
<profile namespace="condor" key="request_disk" >30 GB</profile>
<profile namespace="condor" key="+WantsStashCache" >True</profile>
</site>
By default, the workflow is cloned with requests for at least 5 GB of memory and 30 GB of disk space on OSG compute nodes. If the user is working with an organism with a large reference genome and finds that 5 GB is insufficient, they may change:
<profile namespace="condor" key="request_memory" >5 GB</profile>
to request six gigabytes of RAM:
<profile namespace="condor" key="request_memory" >6 GB</profile>
If the user finds that 5 gigabytes of RAM per job is unnecessary and would like to speed up queue times, they may change:
<profile namespace="condor" key="request_memory" >5 GB</profile>
to request only 3 gigabytes of RAM per job:
<profile namespace="condor" key="request_memory" >3 GB</profile>
This workflow utilizes OASIS software modules that OSG compute nodes can access. Job wrappers in this workflow load these modules to utilize specific versions of software. For example, the following software modules are loaded for all tophat jobs using the 'module load' command:
module load tophat/2.1.1
module load samtools/1.3.1
module load bowtie/2.2.9
module load java/8u25
If the user would like to plug in alternate software, or would like to use a different version of the available software, an osgconnect user support ticket may be submited to have their software of choice installed as an OASIS module.
We have also found that precompiled software packages for linux x86_64 architecture have been stable on OSG compute nodes. The user may utilize these software packages by adding a tar archive of the package to the appropriate task-files directory of the workflow. This will then be transferred as input to the job, which can be unpacked and utilized for the user's task.