Docker and Singularity containers for Polygenic Risk Score analysis

Docker and Singularity containers to perform Polygenic Risk Score (PRS) via PRSice-2 (https://www.prsice.info/) .

Getting Started

In order to run the developed Docker and Singularity containers you need to install Docker (https://docs.docker.com/get-docker/) and Singularity (https://sylabs.io/guides/3.6/user-guide/quick_start.html?highlight=install), respectively. You can run each container separately.

Note that you can try different analysis other than the one presented here as introductive example. For more detailed information related to Prsice-2 analysis, please have a look for [1] (https://www.prsice.info/quick_start/) and [2] (https://www.intro-statistical-genetics.com/data-code ) (ch. 10). Keep in mind that, here we are using "PRSice_linux" instead of "PRSice" hence you may need to change this command if you want to use examples in these references.

For Docker Container

1. Pull the docker container from Docker Hub

docker pull bayramalex/imageprsice

2. You can run the pulled containers in different ways based on the purpose

Passive mode

docker run  bayramalex/imageprsice <your PRSice2 analysis>

Example:

 docker run  bayramalex/imageprsice  Rscript PRSice.R --dir . \
    --prsice ./PRSice_linux \
    --base BMI.txt \
    --target 1kg_hm3_qc \
    --snp MarkerName \
    --A1 A1 \
    --A2 A2 \
    --stat Beta \
    --pvalue Pval \
    --pheno-file BMI_pheno.txt \
    --bar-levels 1 \
    --fastscore \
    --binary-target F \
    --out BMI_score_all
    

Interactive mode

docker run  -it bayramalex/imageprsice 

Once you run the code, you have entered inside the container (indicated by '#' sign), then you can run any anaylsis without typing docker commands as:

Example:

  Rscript PRSice.R --dir . \
    --prsice ./PRSice_linux \
    --base BMI.txt \
    --target 1kg_hm3_qc \
    --snp MarkerName \
    --A1 A1 \
    --A2 A2 \
    --stat Beta \
    --pvalue Pval \
    --pheno-file BMI_pheno.txt \
    --bar-levels 1 \
    --fastscore \
    --binary-target F \
    --out BMI_score_all
    

Mounting your data to container

It is desired to do analysis with your own data. At this point it is required to mount a path to the container and do analysis on the data in this path.

docker run  -it -v  /your/local/path:/INPUT bayramalex/imageprsice 

For example

docker run  -it -v  /Users/bayramakdeniz/mydataset:/INPUT bayramalex/imageprsice 

Once you run this, you will observe an interactive container with an additional INPUT folder which is actually /your/local/path. Now you can do your analysis with your data by indicating the place of it by "INPUT/ " such as:

Rscript PRSice.R --dir . \
    --prsice ./PRSice_linux \
    --base INPUT/BMI.txt \
    --target INPUT/1kg_hm3_qc \
    --thread 1 \
    --snp MarkerName \
    --A1 A1 \
    --A2 A2 \
    --stat Beta \
    --pvalue Pval \
    --bar-levels 5e-08,5e-07,5e-06,5e-05,5e-04,5e-03,5e-02,5e-01 \
    --fastscore \
    --all-score \
    --no-regress \
    --binary-target F \
    --out INPUT/BMIscore_thresholds  

Keep in mind that, the generated outputs after analysis can be both found in INPUT/ and /your/local/path.

It is also possible to run it with passive mode as:

docker run   -v  /your/local/path:/INPUT bayramalex/imageprsice  Rscript PRSice.R --dir . \
   --prsice ./PRSice_linux \
   --base INPUT/BMI.txt \
   --target INPUT/1kg_hm3_qc \
   --thread 1 \
   --snp MarkerName \
   --A1 A1 \
   --A2 A2 \
   --stat Beta \
   --pvalue Pval \
   --bar-levels 5e-08,5e-07,5e-06,5e-05,5e-04,5e-03,5e-02,5e-01 \
   --fastscore \
   --all-score \
   --no-regress \
   --binary-target F \
   --out INPUT/BMIscore_thresholds  

For Singularity Container

1- Build Singularity image from Docker Hub

singularity build imagename.sif docker://bayramalex/imageprsice:latest

Alternatively you can build it as 'sandbox' to use it in passive mode

singularity build --sandbox imagename/  docker://bayramalex/imageprsice:latest

Here a container folder called imagename is created in the working directory. You can cd to this folder and work as if you are working in the passive mode in Docker.

2- Go to the directory where your data is: cd ~/your/local/path

Copy PRSice.R script to this directory.

Run the container by mounting this directory

singularity exec -B  $(pwd):/INPUT /path/of/the/container/imagename.sif  <your PRSice2 analysis>

For Example

singularity exec -B  $(pwd):/INPUT /home/bayram/PRS/imagename.sif Rscript PRSice.R --dir . \
    --prsice /PRSice_linux \
    --base BMI.txt \
    --target 1kg_hm3_qc \
    --snp MarkerName \
    --A1 A1 \
    --A2 A2 \
    --stat Beta \
    --pvalue Pval \
    --pheno-file BMI_pheno.txt \
    --bar-levels 1 \
    --fastscore \
    --binary-target F \
    --extract BMI_score_all.valid  \
    --out BMI_score_all

Warning: A slight difference in Singularity compared to Docker container is: you need to run as /PRSice_linux instead of ./PRSice_linux

References

[1] Choi SW, and O’Reilly PF. "PRSice-2: Polygenic Risk Score Software for Biobank-Scale Data." GigaScience 8, no. 7 (July 1, 2019). https://doi.org/10.1093/gigascience/giz082.

[2] MILLS, Melinda C.; BARBAN, Nicola; TROPF, Felix C. An Introduction to Statistical Genetic Data Analysis. MIT Press, 2020.