Lazy-load R packages

This significantly speeds up conda installation since no R packages
create dependencies. It further solved some issues with R packages being
located in the global conda pkg folder.

README.md

@@ -13,9 +13,8 @@ To run `wg-blimp` you need a UNIX environment that contains a [Bioconda](http://
 It is advised to install `wg-blimp` through Bioconda. It is also recommended to install `wg-blimp` in a fresh environment, as it has many dependencies that may conflict with other packages, for this you can use:
 
 ```
-conda create -n wg-blimp wg-blimp python=3.6
+conda create -n wg-blimp wg-blimp
 ```
-The requirement `python=3.6` is currently necessary because of otherwise slow conda dependency solving.
 
 ### From source
 You can also install `wg-blimp` from source using

environment.yml

@@ -5,7 +5,7 @@ channels:
 dependencies:
   - bedtools
   - bwa
-  - bwameth=0.2.0
+  - bwameth 0.2.0
   - fastqc
   - git
   - h5py
@@ -14,7 +14,6 @@ dependencies:
   - mosdepth
   - multiqc
   - picard
-  - python >=3.5
   - pysam
   - qualimap
   - r-base

snakemake_wrapper/Snakefile

@@ -41,6 +41,18 @@ rule all:
         mbias_ob_reports +
         mbias_ot_reports
 
+
+### R package installation
+
+rule install_r_packages:
+    output:
+        r_session_info = log_dir + '/r_session_info.log'
+    log:
+        log_dir + '/r_install.log'
+    script:
+        'scripts/packageManagement.R'
+
+
 ### ALIGNMENT
 
 rule find_fqs:
@@ -229,7 +241,8 @@ rule multiqc:
 
 rule methylation_metrics:
     input:
-        bed_graphs = expand(methylation_dir + '/{sample}_CpG.bedGraph', sample = config['samples'])
+        bed_graphs = expand(methylation_dir + '/{sample}_CpG.bedGraph', sample = config['samples']),
+        r_installation = rules.install_r_packages.output
     output:
         methylation_metrics = qc_dir + '/methylation_metrics.csv'
     log:
@@ -260,7 +273,8 @@ rule methyl_dackel:
 
 rule bedgraph_to_methylation_ratio:
     input:
-        bedGraph = methylation_dir + '/{sample}_CpG.bedGraph'
+        bedGraph = methylation_dir + '/{sample}_CpG.bedGraph',
+        r_installation = rules.install_r_packages.output
     output:
         bedGraph = temp(methylation_dir + '/{sample}_CpG_ratio.bedGraph')
     log:
@@ -348,7 +362,8 @@ rule camel_dmr:
 
 rule bsseq:
     input:
-        meth = expand(methylation_dir + '/{sample}_CpG.bedGraph', sample = config['samples'])
+        meth = expand(methylation_dir + '/{sample}_CpG.bedGraph', sample = config['samples']),
+        r_installation = rules.install_r_packages.output
     output:
         rdata = bsseq_dir + '/bsseq.Rdata',
         csv   = bsseq_dir + '/dmrs.csv',
@@ -371,7 +386,8 @@ def expected_tool_files():
 
 rule dmr_combination:
     input:
-        **expected_tool_files()
+        **expected_tool_files(),
+        r_installation = rules.install_r_packages.output
     output:
         csv = dmr_dir + '/combined-dmrs.csv',
         bed = dmr_dir + '/dmr-coverage/combined-dmrs.bed'
@@ -414,7 +430,8 @@ rule dmr_annotation:
         repeat_masker_annotation_file = repeat_masker_annotation_file,
         transcript_start_site_file = transcript_start_site_file,
         coverages = expand(dmr_dir + '/dmr-coverage/{sample}.regions.bed.gz', sample = config['samples']),
-        combined_dmrs = dmr_dir + '/combined-dmrs.csv'
+        combined_dmrs = dmr_dir + '/combined-dmrs.csv',
+        r_installation = rules.install_r_packages.output
     output:
         annotated_dmrs = dmr_dir + '/annotated-dmrs.csv'
     params:
@@ -436,7 +453,8 @@ rule methylseekr:
         cgi_annotation_file = cgi_annotation_file,
         gene_annotation_file = gene_annotation_file,
         repeat_masker_annotation_file = repeat_masker_annotation_file,
-        transcript_start_site_file = transcript_start_site_file
+        transcript_start_site_file = transcript_start_site_file,
+        r_installation = rules.install_r_packages.output
     output:
         pmd_all = segmentation_dir + '/pmd-all.csv',
         umr_lmr_all = segmentation_dir + '/umr-lmr-all.csv',
@@ -467,10 +485,13 @@ rule methylseekr:
 
 rule benchmark_plot:
     input:
-        target_files = target_files
+        target_files = target_files,
+        r_installation = rules.install_r_packages.output
     output:
         runtime_report = runtime_report
     params:
         benchmark_dir = benchmark_dir
+    log:
+        log_dir + '/benchmark.log'
     script:
         'scripts/benchmarkPlot.R'

snakemake_wrapper/scripts/benchmarkPlot.R

@@ -1,3 +1,10 @@
+if (exists("snakemake")) {
+  logFile <- file(snakemake@log[[1]])
+
+  sink(logFile, append = TRUE)
+  sink(logFile, append = TRUE, type = "message")
+}
+
 library(ggplot2)
 library(stringr)
 library(data.table)

snakemake_wrapper/scripts/packageManagement.R

@@ -0,0 +1,69 @@
+if (exists("snakemake")) {
+  logFile <- file(snakemake@log[[1]])
+
+  sink(logFile, append = TRUE)
+  sink(logFile, append = TRUE, type = "message")
+}
+
+
+options(repos = c(CRAN = "https://cran.rstudio.com"))
+
+ensurePackageInstallation <- function (packageName, type = "CRAN") {
+
+  isPackageInstalled <- packageName %in% rownames(installed.packages())
+
+  if (!isPackageInstalled) {
+
+    if (type == "CRAN") {
+
+      install.packages(packageName, quiet = TRUE)
+
+    } else if (type == "bioc") {
+
+      if (!requireNamespace("BiocManager", quietly = TRUE))
+        install.packages("BiocManager")
+      BiocManager::install(packageName)
+
+    }
+
+  }
+
+}
+
+loadOrInstall <- function (packageName, type = "CRAN") {
+
+  ensurePackageInstallation(packageName, type)
+
+  library(packageName, character.only = TRUE)
+}
+
+# install dependencies
+
+cranPackages <- c(
+  "data.table",
+  "KernSmooth",
+  "parallel",
+  "ggplot2",
+  "stringr"
+)
+
+biocPackages <- c(
+  "BSgenome",
+  "bsseq",
+  "GenomicRanges",
+  "MethylSeekR",
+  "rtracklayer"
+)
+
+for (package in cranPackages) {
+  loadOrInstall(package)
+}
+
+for (package in biocPackages) {
+  loadOrInstall(package, type = "bioc")
+}
+
+# prepare shiny app
+ensurePackageInstallation("shiny")
+
+writeLines(capture.output(sessionInfo()), snakemake@output$r_session_info)