This section is specific to setting up and running TRIBES in CSIRO HPC cluster pearcey. It can be used as reference for runnign tribes in other cluster environments.
CSIRO HPC cluster is running 20CPU 128GB nodes with slurm as scheduler and env modulules to support versioning of tools and applications.
snakemake --profile cluster --use-singularity -d /scratch1/szu004/tribes/TFCeu estimate_degree_vs_true
This is CSIRO internal setup, but in the futre will be extened to support non CSIRO environments.
Perform the following installation setps.
- install snakemake>=5.4 with in python 3.6.1
- install tribes.tools R package from sources in R/3.5.0
E.g:
module load python/3.6.1
pip install --user --upgrade 'snakemake>=5.4'
module load R/3.5.0
R --no-save
>> in R shell type
install.packages('R/tribes.tools', repos=NULL, type='source')
>> then agree to create a user library
Configure tribes:
cp setup/pearcey/tribesrc ~/.tribesrc
mkdir -p ~/.config/snakemake/cluster
cp setup/pearcey/cluster.config.yaml ~/.config/snakemake/cluster/config.yaml
ssh to one of the pearcey interactive nodes e.g. pearcey-i1.hpc.csiro.au.
Select working directory for the sample dataset e.g. /flush3/$USER/TFCeu
Copy the example dataset with configuration from /flush2/projects/HB_TB_Share/TRIBES/samples/TFCeu to your working dir.
cp -r /flush2/projects/HB_TB_Share/TRIBES/samples/TFCeu /flush3/$USER/TFCeu
To run TRIBES (estimate relatedness) locally using 4 CPU cores:
./tribes -d /flush3/$USER/TFCeu --cores 4 estimate_degree
To run TRIBES using slurm:
./tribes -d /flush3/$USER/TFCeu --profile cluster estimate_degree
To generate report comparing the estimated relatendess against the reported relations:
./tribes -d /flush3/$USER/TFCeu --profile cluster estimate_degree_vs_true
The results are in the working dir /flush3/$USER/TFCeu:
TF-CEU-15-2_BiSnp_EurAF:0.01_LD_GRM-allchr_IBD.csv- estimated IBD0 and EstimatedDegreeTF-CEU-15-2_BiSnp_EurAF:0.01_LD_GRM-allchr_IBD_RVT.html- comparision agains reported relatedness