[Kraken2] Split database from Kraken2 TheiaCoV task

tasks/taxon_id/contamination/task_kraken2.wdl

@@ -1,84 +1,12 @@
 version 1.0
 
-task kraken2_theiacov {
-  input {
-    File read1
-    File? read2
-    String samplename
-    String kraken2_db = "/kraken2-db"
-    Int cpu = 4
-    Int memory = 8
-    String? target_organism
-    Int disk_size = 100
-    String docker_image = "us-docker.pkg.dev/general-theiagen/staphb/kraken2:2.0.8-beta_hv"
-  }
-  command <<<
-    # date and version control
-    date | tee DATE
-    kraken2 --version | head -n1 | tee VERSION
-    num_reads=$(ls *fastq.gz 2> /dev/nul | wc -l)
-    if ! [ -z ~{read2} ]; then
-      mode="--paired"
-    fi
-    echo $mode
-    kraken2 $mode \
-      --threads ~{cpu} \
-      --db ~{kraken2_db} \
-      ~{read1} ~{read2} \
-      --report ~{samplename}_kraken2_report.txt \
-      --output ~{samplename}.classifiedreads.txt
-
-    # Compress and cleanup
-    gzip ~{samplename}.classifiedreads.txt
-
-    percentage_human=$(grep "Homo sapiens" ~{samplename}_kraken2_report.txt | cut -f 1)
-     # | tee PERCENT_HUMAN
-    percentage_sc2=$(grep "Severe acute respiratory syndrome coronavirus 2" ~{samplename}_kraken2_report.txt | cut -f1 )
-     # | tee PERCENT_COV
-    if [ -z "$percentage_human" ] ; then percentage_human="0" ; fi
-    if [ -z "$percentage_sc2" ] ; then percentage_sc2="0" ; fi
-    echo $percentage_human | tee PERCENT_HUMAN
-    echo $percentage_sc2 | tee PERCENT_SC2
-    # capture target org percentage 
-    if [ ! -z "~{target_organism}" ]; then
-      echo "Target org designated: ~{target_organism}"
-      percent_target_organism=$(grep "~{target_organism}" ~{samplename}_kraken2_report.txt | cut -f1 | head -n1 )
-      if [ -z "$percent_target_organism" ] ; then percent_target_organism="0" ; fi
-    else 
-      percent_target_organism=""
-    fi
-    echo $percent_target_organism | tee PERCENT_TARGET_ORGANISM
-
-  >>>
-  output {
-    String date = read_string("DATE")
-    String version = read_string("VERSION")
-    File kraken_report = "~{samplename}_kraken2_report.txt"
-    Float percent_human = read_float("PERCENT_HUMAN")
-    Float percent_sc2 = read_float("PERCENT_SC2")
-    String percent_target_organism = read_string("PERCENT_TARGET_ORGANISM")
-    String? kraken_target_organism = target_organism
-    File kraken2_classified_report = "~{samplename}.classifiedreads.txt.gz" 
-    String docker = docker_image
-    String database = kraken2_db
-  }
-  runtime {
-    docker: docker_image
-    memory: "~{memory} GB"
-    cpu: cpu
-    disks:  "local-disk " + disk_size + " SSD"
-    disk: disk_size + " GB" # TES
-    preemptible: 0
-    maxRetries: 3
-  }
-}
-
 # standalone version (no default database included)
 task kraken2_standalone {
   input {
     File read1
     File? read2
     File kraken2_db
+    String? target_organism
     String samplename
     String docker = "us-docker.pkg.dev/general-theiagen/staphb/kraken2:2.1.2-no-db"
     String kraken2_args = ""
@@ -129,6 +57,21 @@ task kraken2_standalone {
     if [ -z "$percentage_human" ] ; then percentage_human="0" ; fi
     echo $percentage_human | tee PERCENT_HUMAN
 
+    # Report percentage of SARS-CoV-2 reads
+    percentage_sc2=$(grep "Severe acute respiratory syndrome coronavirus 2" ~{samplename}.report.txt | cut -f1 )
+    if [ -z "$percentage_sc2" ] ; then percentage_sc2="0" ; fi
+    echo $percentage_sc2 | tee PERCENT_SC2
+
+    # capture target org percentage
+    if [ ! -z "~{target_organism}" ]; then
+      echo "Target org designated: ~{target_organism}"
+      percent_target_organism=$(grep "~{target_organism}" ~{samplename}.report.txt | cut -f1 | head -n1 )
+      if [ -z "$percent_target_organism" ] ; then percent_target_organism="0" ; fi
+    else
+      percent_target_organism=""
+    fi
+    echo $percent_target_organism | tee PERCENT_TARGET_ORGANISM
+
     # rename classified and unclassified read files if SE
     if [ -e "~{samplename}.classified#.fastq.gz" ]; then
       mv "~{samplename}.classified#.fastq.gz" ~{samplename}.classified_1.fastq.gz
@@ -141,13 +84,16 @@ task kraken2_standalone {
   output {
     String kraken2_version = read_string("VERSION")
     String kraken2_docker = docker
-    String analysis_date = read_string("DATE")
+    String kraken2_analysis_date = read_string("DATE")
     File kraken2_report = "~{samplename}.report.txt"
     File kraken2_classified_report = "~{samplename}.classifiedreads.txt.gz"
     File kraken2_unclassified_read1 = "~{samplename}.unclassified_1.fastq.gz"
     File? kraken2_unclassified_read2 = "~{samplename}.unclassified_2.fastq.gz"
     File kraken2_classified_read1 = "~{samplename}.classified_1.fastq.gz"
     Float kraken2_percent_human = read_float("PERCENT_HUMAN")
+    Float kraken2_percent_sc2 = read_float("PERCENT_SC2")
+    String kraken2_percent_target_organism = read_string("PERCENT_TARGET_ORGANISM")
+    String? kraken2_target_organism = target_organism
     File? kraken2_classified_read2 = "~{samplename}.classified_2.fastq.gz"
     String kraken2_database = kraken2_db
   }
@@ -221,14 +167,14 @@ task kraken2_parse_classified {
     else 
       percent_target_organism=""
     fi
-    echo $percent_target_organism | tee PERCENT_TARGET_ORG
+    echo $percent_target_organism | tee PERCENT_TARGET_ORGANISM
 
   >>>
   output {
     File kraken_report = "~{samplename}.report_parsed.txt"
     Float percent_human = read_float("PERCENT_HUMAN")
     Float percent_sc2 = read_float("PERCENT_SC2")
-    String percent_target_organism = read_string("PERCENT_TARGET_ORG")
+    String percent_target_organism = read_string("PERCENT_TARGET_ORGANISM")
     String? kraken_target_organism = target_organism
   }
   runtime {

tests/inputs/theiacov/wf_theiacov_clearlabs.json

@@ -3,5 +3,6 @@
     "theiacov_clearlabs.read1": "tests/data/theiacov/fastqs/clearlabs/clearlabs.fastq.gz",
     "theiacov_clearlabs.primer_bed": "tests/data/theiacov/primers/artic-v3.primers.bed",
     "theiacov_clearlabs.reference_genome": "tests/data/theiacov/reference/MN908947.fasta",
-    "theiacov_clearlabs.organism_parameters.gene_locations_bed_file": "tests/inputs/sc2_gene_locations.bed"
+    "theiacov_clearlabs.organism_parameters.gene_locations_bed_file": "tests/inputs/sc2_gene_locations.bed",
+    "theiacov_clearlabs.kraken_db": "tests/data/theiacov/databases/github_kraken2_test_db.tar.gz"
 }

tests/inputs/theiacov/wf_theiacov_illumina_pe.json

@@ -5,5 +5,6 @@
     "theiacov_illumina_pe.primer_bed": "tests/data/theiacov/primers/artic-v3.primers.bed",
     "theiacov_illumina_pe.reference_genome": "tests/data/theiacov/reference/MN908947.fasta",
     "theiacov_illumina_pe.reference_gff": "tests/inputs/completely-empty-for-test.txt",
-    "theiacov_illumina_pe.reference_gene_locations_bed": "tests/inputs/sc2_gene_locations.bed"
+    "theiacov_illumina_pe.reference_gene_locations_bed": "tests/inputs/sc2_gene_locations.bed",
+    "theiacov_illumina_pe.kraken_db": "tests/data/theiacov/databases/github_kraken2_test_db.tar.gz"
 }

tests/inputs/theiacov/wf_theiacov_illumina_se.json

@@ -4,5 +4,6 @@
     "theiacov_illumina_se.primer_bed": "tests/data/theiacov/primers/artic-v3.primers.bed",
     "theiacov_illumina_se.reference_genome": "tests/data/theiacov/reference/MN908947.fasta",
     "theiacov_illumina_se.reference_gff": "tests/inputs/completely-empty-for-test.txt",
-    "theiacov_illumina_se.reference_gene_locations_bed": "tests/inputs/sc2_gene_locations.bed"
+    "theiacov_illumina_se.reference_gene_locations_bed": "tests/inputs/sc2_gene_locations.bed",
+    "theiacov_illumina_se.kraken_db": "tests/data/theiacov/databases/github_kraken2_test_db.tar.gz"
 }

tests/inputs/theiacov/wf_theiacov_ont.json

@@ -3,5 +3,6 @@
     "theiacov_ont.read1": "tests/data/theiacov/fastqs/ont/ont.fastq.gz",
     "theiacov_ont.primer_bed": "tests/data/theiacov/primers/artic-v3.primers.bed",
     "theiacov_ont.reference_genome": "tests/data/theiacov/reference/MN908947.fasta",
-    "theiacov_ont.reference_gene_locations_bed": "tests/inputs/sc2_gene_locations.bed"
+    "theiacov_ont.reference_gene_locations_bed": "tests/inputs/sc2_gene_locations.bed",
+    "theiacov_ont.kraken_db": "tests/data/theiacov/databases/github_kraken2_test_db.tar.gz"
 }

tests/workflows/theiacov/test_wf_theiacov_clearlabs.yml

@@ -149,7 +149,7 @@
     - path: miniwdl_run/call-fastq_scan_raw_reads/work/clearlabs_fastq-scan.json
       md5sum: 869dd2e934c600bba35f30f08e2da7c9
     - path: miniwdl_run/call-kraken2_dehosted/command
-      md5sum: 0f9db3341b5f58fb8d145d6d94222827
+      md5sum: 698b918488ef143452798c6c6e5bd0b7
     - path: miniwdl_run/call-kraken2_dehosted/inputs.json
       contains: ["read1", "samplename"]
     - path: miniwdl_run/call-kraken2_dehosted/outputs.json
@@ -161,18 +161,18 @@
       contains: ["wdl", "theiacov_clearlabs", "kraken2_dehosted", "done"]
     - path: miniwdl_run/call-kraken2_dehosted/work/DATE
     - path: miniwdl_run/call-kraken2_dehosted/work/PERCENT_HUMAN
-      md5sum: 4fd4dcef994592f9865e9bc8807f32f4
+      md5sum: 897316929176464ebc9ad085f31e7284
     - path: miniwdl_run/call-kraken2_dehosted/work/PERCENT_SC2
-      md5sum: 9fc4759d176a0e0d240c418dbaaafeb2
+      md5sum: 86b6b8aa9ad17f169f04c02b0e2bf1b1
     - path: miniwdl_run/call-kraken2_dehosted/work/PERCENT_TARGET_ORGANISM
       md5sum: 68b329da9893e34099c7d8ad5cb9c940
     - path: miniwdl_run/call-kraken2_dehosted/work/VERSION
-      md5sum: 379b99c23325315c502e74614c035e7d
+      md5sum: 37b34738ddce9734ae7857ca4d8ee6d3
     - path: miniwdl_run/call-kraken2_dehosted/work/_miniwdl_inputs/0/clearlabs_R1_dehosted.fastq.gz
-    - path: miniwdl_run/call-kraken2_dehosted/work/clearlabs_kraken2_report.txt
-      md5sum: 35841fa2d77ec202c275b1de548b8d98
+    - path: miniwdl_run/call-kraken2_dehosted/work/clearlabs.report.txt
+      md5sum: b66dbcf8d229c1b6fcfff4dd786068bd
     - path: miniwdl_run/call-kraken2_raw/command
-      md5sum: a9dabf08bff8e183fd792901ce24fc57
+      md5sum: c918fc7f9baa660aff0b4fb9e01d5281
     - path: miniwdl_run/call-kraken2_raw/inputs.json
       contains: ["read1", "samplename"]
     - path: miniwdl_run/call-kraken2_raw/outputs.json
@@ -184,16 +184,16 @@
       contains: ["wdl", "theiacov_clearlabs", "kraken2_raw", "done"]
     - path: miniwdl_run/call-kraken2_raw/work/DATE
     - path: miniwdl_run/call-kraken2_raw/work/PERCENT_HUMAN
-      md5sum: 4fd4dcef994592f9865e9bc8807f32f4
+      md5sum: 897316929176464ebc9ad085f31e7284
     - path: miniwdl_run/call-kraken2_raw/work/PERCENT_SC2
-      md5sum: 9fc4759d176a0e0d240c418dbaaafeb2
+      md5sum: 86b6b8aa9ad17f169f04c02b0e2bf1b1
     - path: miniwdl_run/call-kraken2_raw/work/PERCENT_TARGET_ORGANISM
       md5sum: 68b329da9893e34099c7d8ad5cb9c940
     - path: miniwdl_run/call-kraken2_raw/work/VERSION
-      md5sum: 379b99c23325315c502e74614c035e7d
+      md5sum: 37b34738ddce9734ae7857ca4d8ee6d3
     - path: miniwdl_run/call-kraken2_raw/work/_miniwdl_inputs/0/clearlabs.fastq.gz
-    - path: miniwdl_run/call-kraken2_raw/work/clearlabs_kraken2_report.txt
-      md5sum: 35841fa2d77ec202c275b1de548b8d98
+    - path: miniwdl_run/call-kraken2_raw/work/clearlabs.report.txt
+      md5sum: b66dbcf8d229c1b6fcfff4dd786068bd
     - path: miniwdl_run/call-ncbi_scrub_se/command
       contains: ["read1", "scrubber", "gzip"]
     - path: miniwdl_run/call-ncbi_scrub_se/inputs.json

tests/workflows/theiacov/test_wf_theiacov_illumina_pe.yml

@@ -84,7 +84,7 @@
     - path: miniwdl_run/call-read_QC_trim/call-fastq_scan_raw/work/VERSION
     # kraken2 dehosted
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_dehosted/command
-      md5sum: 2031501aaf268d2987b6dbc3b8b32dfa
+      md5sum: ad6608bb5c8581d17a9d14760917171c
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_dehosted/inputs.json
       contains: ["read1", "read2", "samplename"]
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_dehosted/outputs.json
@@ -97,14 +97,14 @@
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_dehosted/work/PERCENT_HUMAN
       md5sum: 897316929176464ebc9ad085f31e7284
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_dehosted/work/PERCENT_SC2
-      md5sum: 494a4bf9ab740c0a0fab64f670549883
+      md5sum: 6baf8bb11094b9011d8dc34e66743712
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_dehosted/work/PERCENT_TARGET_ORGANISM
       md5sum: 68b329da9893e34099c7d8ad5cb9c940
-    - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_dehosted/work/SRR13687078_kraken2_report.txt
-      md5sum: 2ccc036a9a93b3cf096a5c4dda49a579
+    - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_dehosted/work/SRR13687078.report.txt
+      md5sum: 565954ac2bb6ef427754de3b43430728
     # kraken2 raw
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/command
-      md5sum: a16205bdb8cf133a112c4552e8f67f97
+      md5sum: e457432bb1e150e2d561936577c9c51c
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/inputs.json
       contains: ["read1", "samplename"]
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/outputs.json
@@ -115,13 +115,13 @@
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/task.log
       contains: ["wdl", "theiacov_illumina_pe", "kraken2_theiacov_raw", "done"]
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/PERCENT_HUMAN
-      md5sum: 414f4efa514540a2527a4f27124575f2
+      md5sum: 897316929176464ebc9ad085f31e7284
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/PERCENT_SC2
-      md5sum: 2bf2d20f083d8fa09abf6c25f8970e2e
+      md5sum: cfefab882d84cf0f2a1bde9c19eec318
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/PERCENT_TARGET_ORGANISM
       md5sum: 68b329da9893e34099c7d8ad5cb9c940
-    - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/SRR13687078_kraken2_report.txt
-      md5sum: 3544d9ca35d45093c03cdead46677765
+    - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/SRR13687078.report.txt
+      md5sum: 8ea92e13d401e1c955336edfdcd4f1ba
     # ncbi scrub
     - path: miniwdl_run/call-read_QC_trim/call-ncbi_scrub_pe/command
       md5sum: 8c7ca800fa98305009cfb9116a4b60b8

tests/workflows/theiacov/test_wf_theiacov_illumina_se.yml

@@ -74,7 +74,7 @@
     - path: miniwdl_run/call-read_QC_trim/call-fastq_scan_raw/work/VERSION
     # kraken2
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/command
-      md5sum: ca22e45a62c5c26c4447cdafe75a26ab
+      md5sum: a62a43b447fb706b14d0da2069a3aa3f
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/inputs.json
       contains: ["read1", "samplename"]
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/outputs.json
@@ -85,13 +85,13 @@
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/task.log
       contains: ["wdl", "theiacov_illumina_se", "kraken2_theiacov_raw", "done"]
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/PERCENT_HUMAN
-      md5sum: 1576d5d341223ea9d44b0b8a213bb9da
+      md5sum: 897316929176464ebc9ad085f31e7284
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/PERCENT_SC2
-      md5sum: 7cc2eb659e21f15fa902b11812eae1f6
+      md5sum: adbe14d7547234f3743f80907ed33179
     - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/PERCENT_TARGET_ORGANISM
       md5sum: 68b329da9893e34099c7d8ad5cb9c940
-    - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/ERR6319327_kraken2_report.txt
-      md5sum: 9a089b8920e55c9cc7bc8cd7d18f9a8e
+    - path: miniwdl_run/call-read_QC_trim/call-kraken2_theiacov_raw/work/ERR6319327.report.txt
+      md5sum: cb58af9eb139d109b55ce65d6d2344d6
     # clean read screen
     - path: miniwdl_run/call-clean_check_reads/command
       md5sum: aec6c57452ddff84c325601a780605d2
@@ -285,7 +285,7 @@
     - path: miniwdl_run/call-consensus_qc/work/NUM_ACTG
       md5sum: 646a2709ceea763e25df9658d0cf0897
     - path: miniwdl_run/call-consensus_qc/work/NUM_DEGENERATE
-      md5sum: 897316929176464ebc9ad085f31e7284
+      md5sum: 4fd4dcef994592f9865e9bc8807f32f4
     - path: miniwdl_run/call-consensus_qc/work/NUM_N
       md5sum: 7627554073c1f56b9ead3d42d0b8aca8
     - path: miniwdl_run/call-consensus_qc/work/NUM_TOTAL

tests/workflows/theiaprok/test_wf_theiaprok_illumina_pe.yml

@@ -621,17 +621,17 @@
     - path: miniwdl_run/wdl/tasks/species_typing/multi/task_ts_mlst.wdl
       md5sum: ff8070a06eca94264ad6a7d91cb03bf0
     - path: miniwdl_run/wdl/tasks/task_versioning.wdl
-      md5sum: e834231f74072bd3f40861095de25864
+      md5sum: 72d7f4462417909d85f692615e3a658b
     - path: miniwdl_run/wdl/tasks/taxon_id/task_gambit.wdl
       md5sum: 2aa70eab24868920f6c28843dd3b5613
     - path: miniwdl_run/wdl/tasks/taxon_id/contamination/task_kraken2.wdl
-      md5sum: 0ea83681884800bda1e3c4e116f2b19d
+      md5sum: 14e6b12db05686a32208fd7b1c3dfbf8
     - path: miniwdl_run/wdl/tasks/taxon_id/contamination/task_midas.wdl
       md5sum: 64caaaff5910ac0036e2659434500962
     - path: miniwdl_run/wdl/tasks/utilities/data_export/task_broad_terra_tools.wdl
       md5sum: 4d69a6539b68503af9f3f1c2787ff920
     - path: miniwdl_run/wdl/workflows/theiaprok/wf_theiaprok_illumina_pe.wdl
-      md5sum: 6d9dd969e2144ca23f2a0e101e6b6966
+      md5sum: 5b2a4e5b4d7a6306b49af9238527f791
     - path: miniwdl_run/wdl/workflows/utilities/wf_merlin_magic.wdl
       md5sum: 359bc59b671ef3121f6b4ac589f0c80e
     - path: miniwdl_run/wdl/workflows/utilities/wf_read_QC_trim_pe.wdl

tests/workflows/theiaprok/test_wf_theiaprok_illumina_se.yml

@@ -584,20 +584,20 @@
     - path: miniwdl_run/wdl/tasks/species_typing/multi/task_ts_mlst.wdl
       md5sum: ff8070a06eca94264ad6a7d91cb03bf0
     - path: miniwdl_run/wdl/tasks/task_versioning.wdl
-      md5sum: e834231f74072bd3f40861095de25864
+      md5sum: 72d7f4462417909d85f692615e3a658b
     - path: miniwdl_run/wdl/tasks/taxon_id/task_gambit.wdl
       md5sum: 2aa70eab24868920f6c28843dd3b5613
     - path: miniwdl_run/wdl/tasks/taxon_id/contamination/task_kraken2.wdl
-      md5sum: 0ea83681884800bda1e3c4e116f2b19d
+      md5sum: 14e6b12db05686a32208fd7b1c3dfbf8
     - path: miniwdl_run/wdl/tasks/taxon_id/contamination/task_midas.wdl
       md5sum: 64caaaff5910ac0036e2659434500962
     - path: miniwdl_run/wdl/tasks/utilities/data_export/task_broad_terra_tools.wdl
       md5sum: 4d69a6539b68503af9f3f1c2787ff920
     - path: miniwdl_run/wdl/workflows/theiaprok/wf_theiaprok_illumina_se.wdl
-      md5sum: 5aa25e4fad466f92c96a7c138aca0d20
+      md5sum: 6b2bfd31a28f85af9ee6839603f7e4a6
     - path: miniwdl_run/wdl/workflows/utilities/wf_merlin_magic.wdl
       md5sum: 359bc59b671ef3121f6b4ac589f0c80e
     - path: miniwdl_run/wdl/workflows/utilities/wf_read_QC_trim_se.wdl
-      md5sum: d11bfe33fdd96eab28892be5a01c1c7d
+      md5sum: 3ee31b3204261ed54d48c2c2788bdda9
     - path: miniwdl_run/workflow.log
       contains: ["wdl", "theiaprok_illumina_se", "NOTICE", "done"]