Merge pull request #2 from pinin4fjords/limma_formula_jon

Simplify variable filtering

Author: atrigila

assets/differentialabundance_report.Rmd

@@ -367,7 +367,7 @@ if (is.null(differential_file_suffix)) {
     differential_file_suffix <- paste0(".", params$differential_method, ".results.tsv")
 }
 differential_files <- lapply(contrasts$id, function(d){
-    file.path(params$input_dir, paste0(gsub(' |;', '_', d), '_', params$study_name, differential_file_suffix))
+    file.path(params$input_dir, paste0(gsub(' |;', '_', d), differential_file_suffix))
 })
 differential_names <- paste0(contrasts$id, '_', params$study_name)
 

workflows/differentialabundance.nf

@@ -410,6 +410,7 @@ workflow DIFFERENTIALABUNDANCE {
             }
             contrast.formula = contrast.formula?.trim() ? contrast.formula.trim() : null
             contrast.make_contrasts_str = contrast.make_contrasts_str?.trim() ? contrast.make_contrasts_str.trim() : null
+
             return [meta, contrast, contrast.variable, contrast.reference, contrast.target, contrast.formula, contrast.make_contrasts_str]
         }
         .groupTuple() // [meta, [contrast], [variable], [reference], [target], [formula], [comparison]]
@@ -684,39 +685,29 @@ workflow DIFFERENTIALABUNDANCE {
     // create temporary contrast files with the entries based on the order of the gathered
     // differential results
 
-    // As contrasts having 'formula' and 'comparison' won't have a variable,
-    // and it is needed for "checkListIsSubset()" in `SHINYNGS` make_app_from_files.R
-    // for backwards-compatibility, keep only the channels that have non-empty variable
-
-    // Filter by those channels containing 'variable', regroup
-    ch_contrasts_filtered = ch_contrasts
-        .transpose()
-        .filter { meta, contrast, variable, reference, target, formula, comparison ->
-            variable?.trim()
-        }
-        .groupTuple()
-
-    ch_differential_results_with_variable = ch_differential_results
-        .filter { meta, contrast, results ->
-            contrast.variable?.trim()
-        }
-
     // Create a channel with the differential results and the corresponding map with
     // the contrast entries
-    ch_differential_with_contrast = ch_shinyngs
-        .join( ch_differential_results_with_variable.groupTuple() )   // [meta, [meta with contrast], [differential results]]
-        .join( ch_contrasts_filtered )                           // [meta, [contrast], [variable], [reference], [target], [formula], [comparison]]
+    differential_with_contrast = ch_shinyngs
+        .join( ch_differential_results
+            .filter { meta, contrast, results -> contrast.variable?.trim() }
+            .groupTuple() 
+        )   // [meta, [meta with contrast], [differential results]]
+        .join( ch_contrasts )                           // [meta, [contrast], [variable], [reference], [target], [formula], [comparison]]
         .map { meta, meta_with_contrast, results, contrast, variable, reference, target, formula, comparison ->
             // extract the contrast entries from the meta dynamically
             // in this way we don't need to harcode the contrast keys
-            def contrast_keys = contrast[0].keySet()
-            def contrast_maps = meta_with_contrast.collect { it.subMap(contrast_keys) }
-            [meta, results, contrast_maps]   // [meta, [differential results], [contrast maps]]
+            def paramset_contrast_keys = contrast[0].keySet()
+            def contrast_maps = meta_with_contrast.collect { it.subMap(paramset_contrast_keys) }
+            [meta, meta_with_contrast, results, contrast_maps]
+        }
+        .multiMap { meta, meta_with_contrast, results, contrast_maps ->
+            differential_results: [meta, meta_with_contrast, results]
+            contrast_maps: [meta, contrast_maps]
         }
 
     // Save temporary contrast csv files with the entries ordered by the differential results
-    ch_contrasts_sorted = ch_differential_with_contrast
-        .collectFile { meta, results, contrast_map ->
+    ch_contrasts_sorted = differential_with_contrast.contrast_maps
+        .collectFile { meta, contrast_map ->
             def header = contrast_map[0].keySet().join(',')
             def content = contrast_map.collect { it.values().join(',') }
             def lines = header + '\n' + content.join('\n')
@@ -730,10 +721,10 @@ workflow DIFFERENTIALABUNDANCE {
         }
 
     // Parse input for shinyngs app
-    ch_shinyngs_input = ch_differential_with_contrast
+    ch_shinyngs_input = differential_with_contrast.differential_results
         .join(ch_contrasts_sorted)
         .join(ch_all_matrices)
-        .multiMap { meta, differential_results, contrast_map, contrast_file, samplesheet, features, matrices ->
+        .multiMap { meta, meta_with_contrast, differential_results, contrast_file, samplesheet, features, matrices ->
             matrices: [meta, samplesheet, features, matrices]
             contrasts_and_differential: [meta, contrast_file, differential_results]
             contrast_stats_assay: meta.params.exploratory_assay_names.split(',').findIndexOf { it == meta.params.exploratory_final_assay } + 1
@@ -780,7 +771,7 @@ workflow DIFFERENTIALABUNDANCE {
     // So all the files generated with the same paramset meta will go together to report.
     // Note that the files generated with different contrasts will also be grouped together for the same paramset meta.
 
-    ch_differential_grouped = ch_differential_results
+    ch_differential_grouped = differential_with_contrast.differential_results.transpose()
         .join(ch_differential_model, by:[0,1])
         .groupTuple()                                 // [ meta, [meta with contrast], [differential results], [differential model] ]
         .map { [it[0], it.tail().tail().flatten()] }  // [ meta, [differential results and models] ]
@@ -789,45 +780,13 @@ workflow DIFFERENTIALABUNDANCE {
         .groupTuple()                                 // [ meta, [meta with contrast], [functional results] ]
         .map { [it[0], it.tail().tail().flatten()] }  // [ meta, [functional results] ]
 
-    // If users provide a `report_grouping_variable` then update the contrasts file 'variable' column with that information
-    // Split the CSV once at the start
-    ch_split = ch_validated_contrast
-        .splitCsv(header: true, sep: '\t')
-
-    // If the user provided a grouping variable, fill missing/NA values
-    if (params.report_grouping_variable) {
-        ch_split = ch_split
-            .map { meta, row ->
-                def variable = row.variable?.trim()
-                if (!variable || variable == 'NA') {
-                    row.variable = params.report_grouping_variable
-                }
-                [meta, row]
-            }
-    }
-
-    // Now remove any rows where variable is still empty
-    ch_validated_contrast = ch_split
-        .filter { meta, row ->
-            def variable = row.variable?.trim()
-            variable
-        }
-        .groupTuple()
-        .map { meta, rows ->
-            def header = rows[0].keySet().join('\t')
-            def lines = rows.collect { it.values().join('\t') }
-            def content = ([header] + lines).join('\n')
-            def outFile = file("${workflow.workDir}/${meta.id ?: meta.paramset_name}_contrast_variable_filled.tsv")
-            outFile.text = content
-            [meta, outFile]
-        }
     // Prepare input for report generation
     // Each paramset will generate one markdown report by gathering all the files created with the same paramset
 
     ch_report_input = ch_report_files    // [meta, [report_file, logo_file, css_file, citations_file]]
         .combine(ch_collated_versions)   // [versions file]
         .join(ch_all_matrices)           // [meta, samplesheet, features, [matrices]]
-        .join(ch_validated_contrast)     // [meta, contrast file]
+        .join(ch_contrasts_sorted)       // [meta, contrast file]
         .join(ch_differential_grouped)   // [meta, [differential results and models]]
         .join(ch_functional_grouped, remainder: true) // [meta, [functional results]]
         .map { [it[0], it.tail().flatten().grep()] }  // [meta, [files]]   // note that grep() would remove null files from join with remainder true