LSTrAP
LSTrAP, shot for Large Scale Transcriptome Analysis Pipeline, greatly facilitates the construction of co-expression networks from RNA Seq data. The various tools involved are seamlessly connected and CPU-intensive steps are submitted to a computer cluster automatically.
Workflow
LSTrAP wraps multiple existing tools into a single workflow. To use LSTrAP the following tools need to be installed
Steps in bold are submitted to a cluster. Optional steps can be enabled by adding the flag ‑‑enable‑interpro and/or ‑‑enable‑orthology.
Preparation
LSTrAP is designed to run on an Oracle Grid Engine computer cluster system and requires all external tools to be installed on the compute nodes. The "module load" system is supported. A comprehensive list of all tools necessary can be found here. Instructions to run LSTrAP on other systems are provided below.
Installation
Use git to obtain a copy of the LSTrAP code
git clone https://github.molgen.mpg.de/proost/LSTrAP
Next, move into the directory and copy config.template.ini and data.template.ini
cd LSTrAP
cp config.template.ini config.ini
cp data.template.ini data.ini
Configure config.ini and data.ini using the guidelines below
Configuration of LSTrAP
After copying the templates, config.ini needs to be set up to run on your system. It requires the path to Trimmomatic's jar and the modules where Bowtie, Tophat ... are installed in.
The location of the transcriptome data, the refrence genome and a few per-species options need to be defined in data.ini.
Detailed instruction how to set up both configuration files can be found here
Obtaining and preparing data
Scripts to download and prepare data from the Sequence Read Archive are included in LSTrAP in the folder helper. Furthermore, it is recommended to remove splice variants from the GFF3 files, a script to do that is included there as well. Detailed instructions for each script provided to obtain and prepare data can be found here
Running LSTrAP
Once properly configured for your system and data, LSTrAP can be run using a single simple command (that should be executed on the head node)
./run.py config.ini data.ini
Options to enable InterProScan and/or OrthoFinder or to skip certain steps of the pipeline are included, use the command below for more info
./run.py -h
Quality report
After running LSTrAP a log file (lstrap.log) is written, in which samples which failed a quality measure are reported. Note that no samples are excluded from the final network. In case certain samples need to be excluded from the final network remove the htseq file for the sample you which to exclude and re-run the pipeline skipping all steps prior to building the network.
./run.py config.ini data.ini --skip-interpro --skip-orthology --skip-bowtie-build --skip-trim-fastq --skip-tophat --skip-htseq --skip-qc
More information on how the quality of samples is determined can be found here.
Output
Apart from the output all tools included generate, LSTrAP will generate raw and normalized expression matrices, a co‑expression network and co‑expression clusters.
A detailed overview of files produces, including examples, can be found here.
Helper Scripts
LSTrAP comes with a few additional scripts to assist users to download and process data from the Sequence Read Archive, repeat analyses and the case study reported in the manuscript (Proost et al., under preparation).
Details for each script can be found here
Running LSTrAP on transcriptome data
To use LSTrAP on a de novo assembled transcriptome a little pre-processing is required. Instead of the genome a fasta file containing coding sequences can be used (remove UTRs). Using the helper script fasta_to_gtf.py a gtf file suited for LSTrAP can be generated.
python3 fasta_to_gtf.py /path/to/transcript.cds.fasta > output.gtf
Adapting LSTrAP to other cluster managers
LSTrAP is designed and tested on an Oracle Grid Engine, though with minimal effort it can be adopted to run on PBS and Torque based systems (and likely others). First, in the configuration file, check the qsub parameters (e.g. jobs that require multiple CPUs to run -pe cores 4), that differ between systems are set up properly (the nodes and cores on Torque and PBS need to be set using -l nodes=4:ppn=2 to request 4 nodes with 2 processes per node).
Furthermore the submission script might differ, these are located in ./cluster/templates.py . For PBS based systems some settings need to be included by adding #PBS ....
We strive to get LSTrAP running on as many systems as possible. Do not hesitate to contact us in case you experience difficulties running LSTrAP on your system.
Contact
LSTrAP was developed by Sebastian Proost and Marek Mutwil at the Max-Planck Institute for Molecular Plant Physiology
Acknowledgements and Funding
This work is supported by ERA-CAPS though the EVOREPRO project. The authors would like to thank Andreas Donath for technical support and helpful discussions.
License
LSTrAP is freely available under the MIT License