Biowulf High Performance Computing at the NIH
ARDISS: Automatic Relevance Determination for Imputation of GWAS Summary Statistics

ARDISS is a method to impute missing summary statistics in mixed-ethnicity cohorts through Gaussian Process Regression and automatic relevance determination. ARDISS is trained on an external reference panel and does not require information about allele frequencies of genotypes from the original study.


Important Notes

Interactive job
Interactive jobs should be used for debugging, graphics, or applications that cannot be run as batch jobs.

Allocate an interactive session and run the program. Sample session:

[user@biowulf]$ sinteractive --mem=8g  --gres=gpu:k80:1,lscratch:10 -c4 
[user@cn4213 ~]$ module load ARDISS 
[+] Loading cuDNN/7.5/CUDA-10.0 libraries...
[+] Loading CUDA Toolkit  10.0.130  ...
[+] Loading ARDISS 0.1.3  ...
[user@cn4213 ~]$ ardiss -h 
usage: ardiss [-h] --typed_scores TYPED_SCORES --reference_genotypes
              REFERENCE_GENOTYPES --markers MARKERS [--output OUTPUT]
              [--population POPULATION] [--masked MASKED] [--weights WEIGHTS]
              [-w WINDOWSIZE] [-m MAF] [-v] [-g] [--no_weight_optimization]

Impute Summary Statistics using Automatic Relevance Determination

optional arguments:
  -h, --help            show this help message and exit
  --typed_scores TYPED_SCORES
                        Filename for the typed values.
  --reference_genotypes REFERENCE_GENOTYPES
                        Filename for the reference SNPs values.
  --markers MARKERS     Path to the markers file
  --output OUTPUT       Path to the output file
  --population POPULATION
                        Path to the population file
  --masked MASKED       Path to the masked file
  --weights WEIGHTS     Path to the pre-computed weights file
  -w WINDOWSIZE, --windowsize WINDOWSIZE
                        Window size for the moving window, must be even
                        (default: 100)
  -m MAF, --maf MAF     Minor Allele Frequency used for filtering (default:
  -v, --verbose         Tuns on verbose mode
  -g, --gpu             Turn on GPU usage
                        Skip the ARD weight optimization and impute with
                        uniform weights

Copy sample data to your current folder:
[user@cn4213 ~]$  cp $ARDISS_DATA/* . 
[user@cn4213 ~]$  ls 
genotypes.npy  insomnia_typed_zscores  markers.txt
Run ARDISS on the sample data:
[user@cn4213 ~]$  ardiss --typed_scores insomnia_typed_zscores --reference_genotypes genotypes.npy --markers markers.txt

Output files insomnia_typed_zscores.imputed and insomnia_typed_zscores.imputed.weights.txt will be produced. End the interactive session:
[user@cn4213 ~]$ exit
[user@biowulf ~]$