Update benchmark_load workflow

paulsengroup · Feb 8, 2024 · 2d8da61 · 2d8da61
1 parent ce8b4bd
commit 2d8da61
Show file tree

Hide file tree

Showing 2 changed files with 194 additions and 25 deletions.
diff --git a/benchmarks/configs/benchmark_load.config b/benchmarks/configs/benchmark_load.config
@@ -13,17 +13,21 @@ params {
 
     replicates = 3
     resolutions = [10, 100, 500, 1000, 5000, 10000, 25000, 50000, 100000]
+    cpu_cores = [1, 16]
 }
 
 process {
     container = 'ghcr.io/paulsengroup/2023-hictk-paper/hictk-bench:2.0.0'
-    // withName:hictk_load {
-        // stageInMode = 'copy'
-    // }
+    // stageInMode = 'copy'
+    withName:preprocess_pairs_gz {
+        container = 'ghcr.io/paulsengroup/2023-hictk-paper/hic-tools:3.30.00'
+    }
     withName:cooler_cload {
-        // stageInMode = 'copy'
         container = 'ghcr.io/paulsengroup/2023-hictk-paper/cooler-bench:1.0.0'
     }
+    withName:hictools_pre {
+        container = 'ghcr.io/paulsengroup/2023-hictk-paper/hic-tools:3.30.00'
+    }
     withName:summarize {
         container = 'ghcr.io/paulsengroup/2023-hictk-paper/plotting:1.0.0'
     }

diff --git a/benchmarks/workflows/benchmark_load.nf b/benchmarks/workflows/benchmark_load.nf
@@ -8,38 +8,71 @@ nextflow.enable.dsl=2
 workflow {
     Channel.fromPath(params.pairs_file, checkIfExists: true).set { pairs }
     Channel.of(params.resolutions).flatten().set { resolutions }
+    Channel.of(params.cpu_cores).flatten().set { cpus }
     task_ids = Channel.of((1..params.replicates).toList()).flatten()
 
-    preprocess_pairs(
+    preprocess_pairs_zst(
         pairs
     )
 
-    preprocess_pairs.out.pairs
-        .set { pairs_filtered }
+    preprocess_pairs_gz(
+        pairs
+    )
+
+    preprocess_pairs_zst.out.pairs
+        .set { pairs_filtered_zst }
+    preprocess_pairs_gz.out.pairs
+        .set { pairs_filtered_gz }
+
+    task_ids.combine(pairs_filtered_zst)
+        .combine(resolutions)
+        .set { tasks_cooler }
+
+    task_ids.combine(pairs_filtered_zst)
+        .combine(resolutions)
+        .set { tasks_hictk_cool }
 
-    task_ids.combine(pairs_filtered)
+    task_ids.combine(pairs_filtered_zst)
         .combine(resolutions)
-        .set { tasks }
+        .combine(cpus)
+        .set { tasks_hictk_hic }
 
-    hictk_load(
-        tasks,
+    task_ids.combine(pairs_filtered_gz)
+        .combine(resolutions)
+        .combine(cpus)
+        .set { tasks_hictools }
+
+    hictk_load_cool(
+        tasks_hictk_cool,
+        file(params.chrom_sizes)
+    )
+
+    hictk_load_hic(
+        tasks_hictk_hic,
         file(params.chrom_sizes)
     )
 
     cooler_cload(
-        tasks,
+        tasks_cooler,
+        file(params.chrom_sizes)
+    )
+
+    hictools_pre(
+        tasks_hictools,
         file(params.chrom_sizes)
     )
 
     summarize(
-        hictk_load.out.tsv
+        hictk_load_cool.out.tsv
+            .mix(hictk_load_hic.out.tsv)
             .mix(cooler_cload.out.tsv)
-            .map { it[4] }
+            .mix(hictools_pre.out.tsv)
+            .map { it[5] }
             .collect()
     )
 }
 
-process preprocess_pairs {
+process preprocess_pairs_zst {
     label 'process_high'
 
     input:
@@ -59,8 +92,28 @@ process preprocess_pairs {
         '''
 }
 
-process hictk_load {
-    publishDir "${params.outdir}/hictk/", mode: 'copy'
+process preprocess_pairs_gz {
+    label 'process_high'
+
+    input:
+        path pairs
+
+    output:
+        path "*.pairs.gz", emit: pairs
+
+    shell:
+        outname="${pairs.simpleName}_filtered.pairs.gz"
+        '''
+        set -o pipefail
+
+        zcat '!{pairs}' |
+        grep -P 'chr[\\dXY]+\\s\\d+\\schr[\\dXY]+\\s' |
+        pigz -p '!{task.cpus}' -9 > '!{outname}'
+        '''
+}
+
+process hictk_load_cool {
+    publishDir "${params.outdir}/hictk/cool", mode: 'copy'
 
     cpus 1
     memory 36.GB
@@ -77,17 +130,18 @@ process hictk_load {
     output:
         tuple val(id),
               val("hictk"),
-              val("pairs"),
+              val("cooler"),
               val(resolution),
+              val(task.cpus),
               path("*.tsv"), emit: tsv
 
     shell:
-        outname="${id}__${pairs.simpleName}__hictk__pairs__${resolution}.tsv"
+        outname="${id}__${pairs.simpleName}__hictk__cool__${resolution}__${task.cpus}.tsv"
         '''
         set -o pipefail
 
-        printf 'tool\\tformat\\tresolution\\ttime\\tmemory\\tsize\\n' > '!{outname}'
-        printf 'hictk\\tpairs\\t!{resolution}\\t' >> '!{outname}'
+        printf 'tool\\tformat\\tresolution\\tcpus\\ttime\\tmemory\\tsize\\n' > '!{outname}'
+        printf 'hictk\\tcool\\t!{resolution}\\t!{task.cpus}\\t' >> '!{outname}'
 
         mkdir tmp/
         export TMPDIR="$PWD/tmp"
@@ -110,6 +164,60 @@ process hictk_load {
         '''
 }
 
+process hictk_load_hic {
+    publishDir "${params.outdir}/hictk/hic", mode: 'copy'
+
+    cpus 1
+    memory 36.GB
+    label 'process_long'
+
+    tag "${pairs.fileName}_${resolution}_${id}"
+
+    input:
+        tuple val(id),
+              path(pairs),
+              val(resolution),
+              val(cpus)
+        path chrom_sizes
+
+    output:
+        tuple val(id),
+              val("hictk"),
+              val("hic"),
+              val(resolution),
+              val(cpus),
+              path("*.tsv"), emit: tsv
+
+    shell:
+        outname="${id}__${pairs.simpleName}__hictk__hic__${resolution}__${cpus}.tsv"
+        '''
+        set -o pipefail
+
+        printf 'tool\\tformat\\tresolution\\tcpus\\ttime\\tmemory\\tsize\\n' > '!{outname}'
+        printf 'hictk\\thic\\t!{resolution}\\t!{cpus}\\t' >> '!{outname}'
+
+        mkdir tmp/
+        export TMPDIR="$PWD/tmp"
+
+        command time -f '%e\\t%M'             \\
+                     -o '!{outname}'          \\
+                     -a                       \\
+            hictk load '!{chrom_sizes}'       \\
+                       '!{resolution}'        \\
+                       'out.hic'              \\
+                       --format 4dn           \\
+                       --assume-unsorted      \\
+                       --batch-size 50000000  \\
+                       --verbosity=1          \\
+                < <(zstdcat '!{pairs}')
+
+        truncate -s -1 '!{outname}'  # Remove newline
+
+        printf '\\t%d\\n' "$(du -b out.hic | cut -f 1)" >> '!{outname}'
+        '''
+}
+
+
 process cooler_cload {
     publishDir "${params.outdir}/cooler/", mode: 'copy'
 
@@ -128,17 +236,18 @@ process cooler_cload {
     output:
         tuple val(id),
               val("hictk"),
-              val("pairs"),
+              val("cooler"),
               val(resolution),
+              val(task.cpus),
               path("*.tsv"), emit: tsv
 
     shell:
-        outname="${id}__${pairs.simpleName}__cooler__pairs__${resolution}.tsv"
+        outname="${id}__${pairs.simpleName}__cooler__cooler__${resolution}__${task.cpus}.tsv"
         '''
         set -o pipefail
 
-        printf 'tool\\tformat\\tresolution\\ttime\\tmemory\\tsize\\n' > '!{outname}'
-        printf 'cooler\\tpairs\\t!{resolution}\\t' >> '!{outname}'
+        printf 'tool\\tformat\\tresolution\\tcpus\\ttime\\tmemory\\tsize\\n' > '!{outname}'
+        printf 'cooler\\tcool\\t!{resolution}\\t!{task.cpus}\\t' >> '!{outname}'
 
         command time -f '%e\\t%M'          \\
                      -o '!{outname}'       \\
@@ -159,6 +268,62 @@ process cooler_cload {
         '''
 }
 
+
+process hictools_pre {
+    publishDir "${params.outdir}/hictools/", mode: 'copy'
+
+    memory 650.GB
+    label 'process_very_long'
+
+    tag "${pairs.simpleName}"
+
+    input:
+        tuple val(id),
+              path(pairs),
+              val(resolution),
+              val(cpus)
+        path chrom_sizes
+
+    output:
+        tuple val(id),
+              val("hictools"),
+              val("hic"),
+              val(resolution),
+              val(cpus),
+              path("*.tsv"), emit: tsv
+
+    shell:
+        outname="${id}__${pairs.simpleName}__hictools__hic__${resolution}__${cpus}.tsv"
+        memory_gb=task.memory.toGiga()
+        '''
+        set -o pipefail
+
+        printf 'tool\\tformat\\tresolution\\tcpus\\ttime\\tmemory\\tsize\\n' > '!{outname}'
+        printf 'hictools\\thic\\t!{resolution}\\t!{cpus}\\t' >> '!{outname}'
+
+        mkdir tmp/
+        export TMPDIR="$PWD/tmp"
+
+        command time -f '%e\\t%M'          \\
+                     -o '!{outname}'       \\
+                     -a                    \\
+            run_hic_tools_pre.sh \\
+                '!{pairs}' \\
+                out.hic \\
+                "$TMPDIR" \\
+                "$HICTOOLS_JAR" \\
+                '!{resolution}' \\
+                '!{cpus}' \\
+                !{memory_gb}G
+
+        java -jar -Xmx
+
+        truncate -s -1 '!{outname}'  # Remove newline
+
+        printf '\\t%d\\n' "$(du -b out.hic | cut -f 1)" >> '!{outname}'
+        '''
+}
+
 process summarize {
     publishDir "${params.outdir}/", mode: 'copy'