WGS bioinfo with NUS-HPC/NSCC

This repository includes example files, "MW.sh" and "EM.sh", for submitting tasks to NUS-HPC and NSCC (National Supercomupting Centre Singapore), respectively.

The variant calling protocol generally follows the automated pipeline "snpArcher" (Mirchandani et al., 2023), empowered by snakemake. https://github.com/harvardinformatics/snpArcher

Setup

Create a conda environment for snakemake and snparcher.

Then install snparcher to your "hpctmp" of 500G (request to get maximum 2TB for free).

If still not enough, try NSCC with quota of 100TB space.....

NUS-HPC allows longer walltime, whereas NSCC can allocate more threads and memory for a single run but check your quota of cpu time.

module load miniconda
# module load miniforge3 #(if use NSCC)
conda create --name snparcher
source activate snparcher
conda install -c bioconda snakemake
cd /hpctmp/dbstq/
# cd /scratch/users/nus/dbstq/ #(if use NSCC)
git clone https://github.com/harvardinformatics/snpArcher.git

By the end, you will have a folder named "snpArcher" under your "hpctmp".

Conda environments take space in your home directory (50GB only)!! Please do housekeeping and remove obsolete environments regularly.

We also uploaded example files with Sentieon (https://www.sentieon.com/) using NSCC.

Change the rule files under "/snpArcher/workflow/rules" when necessary (for examples):

1. To aquire final vcf with both invariant and variant site, replace "VARIANT" with "CONFIDENT" for flag "--emit_mode" under "rule sentieon_combine_gvcf" in rule file "sentieon.smk".
2. For chromosomes exceed the length of 512Mbp, replace "bam" with "cram", and "bai" with "crai" across multiple rule files.

Start a project

For each project (normally a species/species complex sharing the same reference genome), create a project folder and copy the "config" folder.

mkdir projectX
mkdir projectX/data
mkdir projectX/results
cp snpArcher/config projectX -r

By the end, in your "hpctmp", you will have the following directory structure:

├── snpArcher
├── project_1/
│   ├── project1.sh
│   ├── config/
│   │   ├── config.yaml
│   │   ├── resources.yaml
│   │   └── samples.csv
│   ├── data
│   │   ├── ref_genome/
│   │   │   ├── ././ref_genome.fna
│   │   ├── sample1_1.fq.gz
│   │   ├── sample1_2.fq.gz
│   │   ├── sample2_1.fq.gz
│   │   ├── sample2_2.fq.gz
│   └── results
└── project_2/
    ├── project2.sh
    ├── config/
    │   ├── config.yaml
    │   ├── resources.yaml
    │   └── samples.csv
    └── data

Upload raw reads and reference genome to the "projectX/data" folder.

Edit the sample file (samples.csv) and configuration files (config.yaml and resources.yaml) in "projectX/config".

Create a shell file to submit job. For detailed instruction, please see: https://snparcher.readthedocs.io/en/latest/executing.html#cluster-execution

Please note that the pipeline may need access to the Internet, so use cluster "gold" or "tiger2" in the hpc.

Use command

gstat

to check the availability of the clusters.

Finally, when everything is set, run:

qsub projectX.sh

References

Mirchandani, C. D., Shultz, A. J., Thomas, G. W., Smith, S. J., Baylis, M., Arnold, B., ... & Sackton, T. B. (2023). A fast, reproducible, high-throughput variant calling workflow for population genomics. Molecular Biology and Evolution, msad270.

Acknowledgements

“We would like to acknowledge that computational work involved in this research work is partially/fully supported by NUS IT’s Research Computing group using grant numbers NUSREC-HPC-00001”.

Contact author

[email protected]