Merge pull request #34 from sanger-tol/dp24_barcodes

Dp24 barcodes

assets/github_testing/test.yaml

@@ -1,6 +1,7 @@
 assembly_path: /home/runner/work/ascc/ascc/asccTinyTest/assembly/Pyoeliiyoelii17XNL_assembly.fa
 assembly_title: asccTinyTest
-pacbio_multiplexing_barcode_names: ""
+pacbio_barcodes: /home/runner/work/ascc/ascc/assets/pacbio_adaptors.fa
+pacbio_multiplexing_barcode_names: "bc1008,bc1009"
 pacbio_reads_path: /home/runner/work/ascc/ascc/asccTinyTest/pacbio/
 sci_name: "Plasmodium yoelii yoelii 17XNL"
 taxid: 352914

assets/pacbio_adaptors.fa

@@ -1,32 +1,224 @@
->bc1001_BAK8A_OA
+>bc1001
 CACATATCAGAGTGCGT
->bc1002_BAK8A_OA
+>bc1002
 ACACACAGACTGTGAGT
->bc1003_BAK8A_OA
+>bc1003
 ACACATCTCGTGAGAGT
->bc1008_BAK8A_OA
+>bc1008
 ACAGTCGAGCGCTGCGT
->bc1009_BAK8A_OA
+>bc1009
 ACACACGCGAGACAGAT
->bc1010_BAK8A_OA
+>bc1010
 ACGCGCTATCTCAGAGT
->bc1011_BAK8A_OA
+>bc1011
 CTATACGTATATCTATT
->bc1012_BAK8A_OA
+>bc1012
 ACACTAGATCGCGTGTT
->bc1015_BAK8B_OA
+>bc1015
 CGCATGACACGTGTGTT
->bc1016_BAK8B_OA
+>bc1016
 CATAGAGAGATAGTATT
->bc1017_BAK8B_OA
+>bc1017
 CACACGCGCGCTATATT
->bc1018_BAK8B_OA
+>bc1018
 TCACGTGCTCACTGTGT
->bc1019_BAK8B_OA
+>bc1019
 ACACACTCTATCAGATT
->bc1020_BAK8B_OA
+>bc1020
 CACGACACGACGATGTT
->bc1021_BAK8B_OA
+>bc1021
 CTATACATAGTGATGTT
->bc1022_BAK8B_OA
+>bc1022
 CACTCACGTGTGATATT
+>bc2001
+ATCGTGCGACGAGTAT
+>bc2002
+TGCATGTCATGAGTAT
+>bc2003
+ACGAGTGCTCGAGTAT
+>bc2004
+TGCAGTGCTCGAGTAT
+>bc2005
+TGACTCGATCGAGTAT
+>bc2006
+CATGCGATCTGAGTAT
+>bc2007
+ACTAGCATCTGAGTAT
+>bc2008
+ACGCTAGTCTGAGTAT
+>bc2009
+CGATCGCACTGAGTAT
+>bc2010
+TACGTAGTATGAGTAT
+>bc2011
+CTGACAGTACGAGTAT
+>bc2012
+TCGTACTACTGAGTAT
+>bc2013
+CTGCGTAGACGAGTAT
+>bc2014
+ATACATGCACGAGTAT
+>bc2015
+CGACATAGATGAGTAT
+>bc2016
+ATCTGCACGTGAGTAT
+>bc2017
+CTATGATAGCGAGTAT
+>bc2018
+CGATCAGTGCGAGTAT
+>bc2019
+CGTCATAGTCGAGTAT
+>bc2020
+ACTATGCGTCGAGTAT
+>bc2021
+CGTACATGCTGAGTAT
+>bc2022
+TCATCGACGTGAGTAT
+>bc2023
+TCGCATGACTGAGTAT
+>bc2024
+CATGATCGACGAGTAT
+>bc2025
+ACGCACGTACGAGTAT
+>bc2026
+CAGTAGCGTCGAGTAT
+>bc2027
+TGACTGTAGCGAGTAT
+>bc2028
+ACTGCAGCACGAGTAT
+>bc2029
+TAGCAGTATCGAGTAT
+>bc2030
+CATACAGCATGAGTAT
+>bc2031
+ATAGCGTACTGAGTAT
+>bc2032
+ATAGACGAGTGAGTAT
+>bc2033
+CGACTCGTATGAGTAT
+>bc2034
+TACTAGTGACGAGTAT
+>bc2035
+CAGCTGACATGAGTAT
+>bc2036
+ACGTCGCTGCGAGTAT
+>bc2037
+CAGTATGAGCGAGTAT
+>bc2038
+TCACGACGACGAGTAT
+>bc2039
+CATGTATGTCGAGTAT
+>bc2040
+TGCTGCGACTGAGTAT
+>bc2041
+TATGATCACTGAGTAT
+>bc2042
+TCTGCACTGCGAGTAT
+>bc2043
+ACGATGACGTGAGTAT
+>bc2044
+CGATGATGCTGAGTAT
+>bc2045
+TACGACAGTCGAGTAT
+>bc2046
+ATAGCATGTCGAGTAT
+>bc2047
+CATAGTACTCGAGTAT
+>bc2048
+TGATGCTAGTGAGTAT
+>bc2049
+TAGTCTGCGTGAGTAT
+>bc2050
+CTCATCTATCGAGTAT
+>bc2051
+TGCATACTGCGAGTAT
+>bc2052
+CAGACTAGTCGAGTAT
+>bc2053
+ATCGTGATCTGAGTAT
+>bc2054
+CTGCGATCACGAGTAT
+>bc2055
+CTCAGCATACGAGTAT
+>bc2056
+TCGCAGCGTCGAGTAT
+>bc2057
+TAGCACGCATGAGTAT
+>bc2058
+TACTGACGCTGAGTAT
+>bc2059
+ATCTGACTATGAGTAT
+>bc2060
+ATACGAGCTCGAGTAT
+>bc2061
+CGAGCACGCTGAGTAT
+>bc2062
+TCTGCGTATCGAGTAT
+>bc2063
+TCTGCATCATGAGTAT
+>bc2064
+TGCGTGATGCGAGTAT
+>bc2065
+TGAGCTATGCGAGTAT
+>bc2066
+CTGTCGTAGTGAGTAT
+>bc2067
+ATCGATGCATGAGTAT
+>bc2068
+ACTACGTGATGAGTAT
+>bc2069
+TCTATGACATGAGTAT
+>bc2070
+TACTGCTCACGAGTAT
+>bc2071
+CGAGTCTAGCGAGTAT
+>bc2072
+TATCAGTAGTGAGTAT
+>bc2073
+ATCACTAGTCGAGTAT
+>bc2074
+TATCACGACTGAGTAT
+>bc2075
+CTCGTCAGATGAGTAT
+>bc2076
+CAGCAGTGACGAGTAT
+>bc2077
+TGCGACGTGCGAGTAT
+>bc2078
+CTCACTGAGTGAGTAT
+>bc2079
+CACTGAGCGTGAGTAT
+>bc2080
+CAGCGTCTACGAGTAT
+>bc2081
+CTACTATGTCGAGTAT
+>bc2082
+ATGTACAGACGAGTAT
+>bc2083
+ACTCATCAGTGAGTAT
+>bc2084
+CTGAGCACTCGAGTAT
+>bc2085
+ATCATCTACTGAGTAT
+>bc2086
+TACATGCGATGAGTAT
+>bc2087
+TCGCTGTCACGAGTAT
+>bc2088
+ACGCTCATGCGAGTAT
+>bc2089
+TACTAGCAGCGAGTAT
+>bc2090
+CGTAGCAGATGAGTAT
+>bc2091
+CGTGCTCGTCGAGTAT
+>bc2092
+ACAGCTGTACGAGTAT
+>bc2093
+TCGATGCTACGAGTAT
+>bc2094
+TAGATACAGCGAGTAT
+>bc2095
+CTACTCATACGAGTAT
+>bc2096
+ATGTACTAGTGAGTAT

assets/test.yaml

@@ -1,6 +1,7 @@
 assembly_path: /lustre/scratch123/tol/resources/treeval/treeval-testdata/asccTinyTest/assembly/Pyoeliiyoelii17XNL_assembly.fa
 assembly_title: asccTinyTest
-pacbio_multiplexing_barcode_names: something
+pacbio_barcodes: /nfs/treeoflife-01/teams/tola/users/dp24/ascc/assets/pacbio_adaptors.fa
+pacbio_multiplexing_barcode_names: "bc1008,bc1009"
 pacbio_reads_path: /lustre/scratch123/tol/resources/treeval/treeval-testdata/asccTinyTest/pacbio/
 sci_name: "Plasmodium yoelii yoelii 17XNL"
 taxid: 352914

bin/pacbio_barcode_check.py

@@ -1,5 +1,14 @@
+#!/usr/bin/env python3
+
 """
-Notes: Forces sys.exit(1) to kill pipeline
+Pacbio Barcode Check
+------------------------
+Looks for Pacbio barcodes in ref and data.
+
+If any User-supplied barcodes arn't in data then pipeline dies.
+If no barcodes are found then pipeline carries on
+
+Based on a standard operating procedure developed by James Torrance
 
 Originally written by Eerik Aunin @eeaunin
 
@@ -21,7 +30,9 @@ def detect_barcodes_from_read_file_names(barcodes_fasta_path, pacbio_read_files)
     barcodes_fasta_data = gpf.l(barcodes_fasta_path)
     barcode_names = [n.split(">")[1] for n in barcodes_fasta_data if n.startswith(">")]
     if len(barcode_names) == 0:
-        print("NO BARCODES, KILL PIPELINE")
+        sys.stderr.write(
+            f"Failed to read PacBio multiplexing barcode names from the specified file {barcodes_fasta_data}\n"
+        )
         sys.exit(1)
     detected_barcodes = list()
     for barcode_name in barcode_names:
@@ -40,44 +51,52 @@ def check_if_barcodes_exist_in_barcodes_fasta(barcodes_list, barcodes_fasta_path
     barcode_names_in_fasta = [n.split(">")[1] for n in barcodes_fasta_data if n.startswith(">")]
     for barcode in barcodes_list:
         if barcode not in barcode_names_in_fasta:
-            # sys.stderr.write(f"The PacBio multiplexing barcode ({barcode}) was not found in the barcode sequences file ({barcodes_fasta_path})\n")
-            print("NO BARCODES, KILL PIPELINE")
+            sys.stderr.write(
+                f"The PacBio multiplexing barcode ({barcode}) was not found in the barcode sequences file ({barcodes_fasta_path})\n"
+            )
             sys.exit(1)
 
+    # If this print statement is reached, all user-supplied codes are present.
+    print("The query barcodes exist in the barcodes database file")
+
 
 def main(barcodes_fasta_path, pacbio_read_files, pacbio_multiplexing_barcode_names):
     pacbio_read_files = pacbio_read_files.split(",")
 
     barcodes_list = []
-    if pacbio_multiplexing_barcode_names != "NA":
-        barcodes_list = pacbio_multiplexing_barcode_names.split(",")
+    if len(pacbio_multiplexing_barcode_names) > 0:
+        barcodes_list = pacbio_multiplexing_barcode_names.strip("[").strip("]").split(",")
 
-    current_script_dir = os.path.dirname(sys.argv[0])
-
-    if barcodes_fasta_path is None:
-        barcodes_fasta_path = f"{current_script_dir}/third_party_files/pacbio_barcode_screen/pacbio_adaptors.fa"
-    else:
-        if os.path.isfile(barcodes_fasta_path) is False:
-            print("NO BARCODES, KILL PIPELINE")
-            sys.exit(1)
+    if os.path.isfile(barcodes_fasta_path) is False:
+        sys.stderr.write(
+            "FASTA file with PacBio multiplexing barcode sequences ({barcodes_fasta_path}) was not found\n"
+        )
+        sys.exit(1)
 
-    if barcodes_list == []:
+    if len(barcodes_list) == 0:
         barcodes_list = detect_barcodes_from_read_file_names(barcodes_fasta_path, pacbio_read_files)
+
+    # Here script should break successfully
     if len(barcodes_list) == 0:
-        print("NO BARCODES, KILL PIPELINE")
-        sys.exit(1)
+        sys.stderr.write(
+            "Skipping the PacBio barcodes check, as no barcodes were specified by the user and no barcodes were found in PacBio read file names\n"
+        )
+        print(
+            "Skipping the PacBio barcodes check, as no barcodes were specified by the user and no barcodes were found in PacBio read file names\n"
+        )
+        sys.exit(0)
 
-    check_if_barcodes_exist_in_barcodes_fasta(
-        barcodes_list, barcodes_fasta_path
-    )  # This is a TRUE | FALSE check, if FALSE kill pipeline.
-    print("BARCODES FOUND!")
+    check_if_barcodes_exist_in_barcodes_fasta(barcodes_list, barcodes_fasta_path)
 
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description=__doc__)
-    parser.add_argument("barcode_fasta", type=str, help="Pacbio Barcode FASTA file")
-    parser.add_argument("pacbio_reads", type=str, help="Pacbio Read FASTA.gz files")
-    parser.add_argument("multiplex_name", type=str, help="Pacbio Multiplex Barcode Name")
+    parser.add_argument("-b", "--barcode_fasta", type=str, help="Pacbio Barcode FASTA file")
+
+    parser.add_argument("-p", "--pacbio_reads", type=str, help="Pacbio Read FASTA.gz files")
+
+    parser.add_argument("-m", "--multiplex_name", type=str, help="Pacbio Multiplex Barcode Name")
+
     parser.add_argument("-v", action="version", version="1.0")
     args = parser.parse_args()
     main(args.barcode_fasta, args.pacbio_reads, args.multiplex_name)

conf/modules.config

@@ -26,6 +26,10 @@ process {
         ext.args = 'nucleotide'
     }
 
+    withName: BLAST_MAKEBLASTDB {
+        ext.args = { "-dbtype nucl" }
+    }
+
     withName: BLAST_BLASTN {
         ext.args = { "-outfmt '6 qseqid staxids bitscore std' -max_target_seqs 10 -max_hsps 1 -evalue 1e-25 -dust yes -lcase_masking" }
     }