LDpred2.R: redirect temporary data during runs (#204)

* `LDpred2.R`: redirect temporary data
Fixes #203

* Check if tempdir exist before creating tempfile

* updated changelog

CHANGELOG.md

@@ -38,6 +38,13 @@ If MD5 sum is not listed for a certain release then it means that the container
 ### Misc
 
 * Miscellaneous goes here
+
+## [1.3.9] - 2023-10-17
+
+### Added
+
+* User-set directory option for temporary files during LDpred2 runs, by default `base::tempdir()`
+
 ## [1.3.8] - 2023-10-17
 
 ### Fixed
@@ -50,6 +57,10 @@ If MD5 sum is not listed for a certain release then it means that the container
 
 * Added a feature to read and convert BGEN (.bgen) files to ``scripts/pgs/LDpred2/createBackingFile.R``
 
+## [1.3.7] - 2023-10-12
+
+* User-set directory for temporary files during LDpred2 runs, by default `base::tempdir()`
+
 ## [1.3.6] - 2023-08-17
 
 ### Fixed

scripts/pgs/LDpred2/README.md

@@ -43,7 +43,7 @@ usage: ldpred2.R [--] [--help] [--out-merge] [--geno-impute-zero]
        N-CONTROLS] [--name-score NAME-SCORE] [--hyper-p-length
        HYPER-P-LENGTH] [--hyper-p-max HYPER-P-MAX] [--ldpred-mode
        LDPRED-MODE] [--cores CORES] [--set-seed SET-SEED]
-       [--genomic-build GENOMIC-BUILD]
+       [--genomic-build GENOMIC-BUILD] [--tmp-dir TMP-DIR]
 
 Calculate polygenic scores using ldpred2
 
@@ -336,3 +336,20 @@ As above, ``<path/to/containers`` should point to the cloned ``containers`` repo
 Entries like ``--partition=normal`` may also be adapted for different HPC resources.
 Then, the job can be submitted to the queue by issuing ``sbatch run_ldpred2_slurm.job``.
 The status of running jobs can usually be enquired by issuing ``squeue -u $USER``.
+
+
+### Redirect temporary file output
+
+By default, the LDpred2.R script will put large file(s) in the system temporary directory (using `base::tempdir()`).
+For use on HPC resources, use of the designated `$SCRATCH`, `$LOCALTMP`, or `$TMPDIR` directories is recommended to avoid
+filling up the system temporary directory.
+
+One can redirect the temporary file output by setting the `TMPDIR` environment variable to a mounted directory on the HPC resource, 
+by incorporating the following lines to the job script:
+
+```
+export SINGULARITY_BIND=$REFERENCE:/REF,${LDPRED2_REF}:/ldpred2_ref,$SCRATCH:/scratch
+export SINGULARITYENV_TMPDIR=/scratch
+```
+
+Otherwise, the location of temporary files can be specified by the `--tmp-dir` argument to the `ldpred2.R` script.

scripts/pgs/LDpred2/ldpred2.R

@@ -54,6 +54,7 @@ par <- add_argument(par, "--cores", help="Number of CPU cores to use, otherwise
 par <- add_argument(par, '--set-seed', help="Set a seed for reproducibility", nargs=1)
 par <- add_argument(par, "--merge-by-rsid", help="Merge using rsid (the default is to merge by chr:bp:a1:a2 codes).", flag=TRUE)
 par <- add_argument(par, "--genomic-build", help="Genomic build to use. Either hg19, hg18 or hg38", default="hg19", nargs=1)
+par <- add_argument(par, "--tmp-dir", help="Directory to store temporary files. Default is output of base::tempdir()", default=tempdir())
 
 parsed <- parse_args(par)
 
@@ -135,6 +136,9 @@ if (fileOutputMerge) {
   verifyScoreOutputFile(fileOutput, nameScore, fileOutputMergeIDs)
 }
 
+# check if tmp dir exists
+if (!file.exists(parsed$tmp_dir)) stop("Temporary directory", parsed$tmp_dir, "does not exist") 
+
 cat('Loading backingfile:', fileGeno ,'\n')
 obj.bigSNP <- snp_attach(fileGeno)
 
@@ -249,7 +253,7 @@ drops <- c("_NUM_ID_.ss", "rsid.ss", 'block_id', 'pos_hg18', 'pos_hg38')
 df_beta <- df_beta[ , !(names(df_beta) %in% drops)]  
 
 cat('\n### Loading LD reference from ', fileLD, '\n')
-tmp <- tempfile(tmpdir = "tmp-data")
+tmp_file <- tempfile(tmpdir=parsed$tmp_dir)
 ld_size <- 0; corr <- NULL
 for (chr in chr2use) {
   ## indices in 'df_beta' corresponding to a particular 'chr'
@@ -269,7 +273,7 @@ for (chr in chr2use) {
   corr_chr <- readRDS(fileLD_chr)[ind.chr3, ind.chr3]
 
   if (is.null(corr)) {
-    corr <- as_SFBM(corr_chr, tmp, compact = TRUE)
+    corr <- as_SFBM(corr_chr, tmp_file, compact = TRUE)
   } else {
     corr$add_columns(corr_chr, nrow(corr))
   }
@@ -335,4 +339,4 @@ if (fileOutputMerge) cat('Merging by', paste0(fileOutputMergeIDs, collapse=', ')
 writeScore(obj.bigSNP$fam, fileOutput, nameScore, fileOutputMerge, fileOutputMergeIDs)
 cat('Scores written to', fileOutput, '\n')
 # Drop temporary file
-fileRemoved <- file.remove(paste0(tmp, '.sbk'))
+fileRemoved <- file.remove(paste0(tmp_file, '.sbk'))

version/version.py

@@ -2,7 +2,7 @@
 _MINOR = "3"
 # On main and in a nightly release the patch should be one ahead of the last
 # released build.
-_PATCH = "8"
+_PATCH = "9"
 # This is mainly for nightly builds which have the suffix ".dev$DATE". See
 # https://semver.org/#is-v123-a-semantic-version for the semantics.
 _SUFFIX = ""