treesak 1.53.3__py3-none-any.whl

TreeSAK/FigTree.py

@@ -0,0 +1,34 @@
+import os
+import argparse
+from ete3 import Tree
+
+
+FigTree_usage = '''
+====================== FigTree example commands ======================
+
+TreeSAK FigTree -h
+
+======================================================================
+'''
+
+
+def FigTree(args):
+
+    input_txt_file  = args['i']
+    op_txt          = args['o']
+
+    if os.path.isfile(input_txt_file) is False:
+        print('Metadata file not found, program exited!')
+        exit()
+
+
+if __name__ == '__main__':
+
+    FigTree_parser = argparse.ArgumentParser(usage=FigTree_usage)
+    FigTree_parser.add_argument('-i',     required=True,                         help='input metadata')
+    FigTree_parser.add_argument('-tree',  required=False, default=None,          help='gene id, in tree file')
+    FigTree_parser.add_argument('-txt',   required=False, default=None,          help='gene id, in txt file')
+    FigTree_parser.add_argument('-o',     required=True,                         help='output metadata')
+    FigTree_parser.add_argument('-na',    required=False, action='store_true',   help='include leaves with na values')
+    args = vars(FigTree_parser.parse_args())
+    FigTree(args)

TreeSAK/GTDB_tree.py

@@ -0,0 +1,76 @@
+import os
+import argparse
+
+
+GTDB_tree_usage = '''
+======================== GTDB_tree example command ========================
+
+export GTDBTK_DATA_PATH=/scratch/PI/boqianpy/Database/gtdb_r220/release220
+TreeSAK GTDB_tree -p Demo -i gnm_folder -x fa -t 12
+
+# This is a wrapper for the following commands
+gtdbtk identify --genome_dir gnm_folder -x fa --out_dir op_dir --cpus 12
+gtdbtk align --identify_dir Demo_op_dir --out_dir op_dir --cpus 12
+gtdbtk infer --msa_file Demo_op_dir/align/gtdbtk.bac120.user_msa.fasta.gz --out_dir op_dir --cpus 12 --prefix Demo_bac120
+gtdbtk infer --msa_file Demo_op_dir/align/gtdbtk.ar53.user_msa.fasta.gz --out_dir op_dir --cpus 12 --prefix Demo_ar53
+
+===========================================================================
+'''
+
+def GTDB_tree(args):
+
+    input_gnm_dir       = args['i']
+    output_prefix       = args['p']
+    file_extension      = args['x']
+    num_threads         = args['t']
+
+    output_dir          = '%s_GTDB_tree'                                                            % output_prefix
+    msa_bac120_gz       = '%s/align/gtdbtk.bac120.user_msa.fasta.gz'                                % output_dir
+    msa_bac120          = '%s/align/gtdbtk.bac120.user_msa.fasta'                                   % output_dir
+    msa_ar53_gz         = '%s/align/gtdbtk.ar53.user_msa.fasta.gz'                                  % output_dir
+    msa_ar53            = '%s/align/gtdbtk.ar53.user_msa.fasta'                                     % output_dir
+
+    cmd_identify        = 'gtdbtk identify --genome_dir %s -x %s --out_dir %s --cpus %s --write_single_copy_genes'  % (input_gnm_dir, file_extension, output_dir, num_threads)
+    cmd_align           = 'gtdbtk align --identify_dir %s --out_dir %s --cpus %s'                                   % (output_dir, output_dir, num_threads)
+    cmd_gunzip_bac120   = 'gunzip %s'                                                                               % msa_bac120_gz
+    cmd_gunzip_ar53     = 'gunzip %s'                                                                               % msa_ar53_gz
+    cmd_infer_bac120    = 'gtdbtk infer --msa_file %s --out_dir %s --cpus %s --prefix %s_bac120'                    % (msa_bac120, output_dir, num_threads, output_prefix)
+    cmd_infer_ar53      = 'gtdbtk infer --msa_file %s --out_dir %s --cpus %s --prefix %s_ar53'                      % (msa_ar53, output_dir, num_threads, output_prefix)
+
+    print(cmd_identify)
+    os.system(cmd_identify)
+    print(cmd_align)
+    os.system(cmd_align)
+
+    if os.path.isfile(msa_bac120_gz):
+        print(cmd_gunzip_bac120)
+        os.system(cmd_gunzip_bac120)
+        print(cmd_infer_bac120)
+        os.system(cmd_infer_bac120)
+
+    if os.path.isfile(msa_ar53_gz):
+        print(cmd_gunzip_ar53)
+        os.system(cmd_gunzip_ar53)
+        print(cmd_infer_ar53)
+        os.system(cmd_infer_ar53)
+
+    inferred_bac120_tree = '%s/%s_bac120.unrooted.tree' % (output_dir, output_prefix)
+    inferred_ar53_tree   = '%s/%s_ar53.unrooted.tree'   % (output_dir, output_prefix)
+
+    if os.path.isfile(inferred_bac120_tree):
+        print('Inferred bacterial tree:\t%s' % inferred_bac120_tree)
+    if os.path.isfile(inferred_ar53_tree):
+        print('Inferred archaeal tree:\t%s' % inferred_ar53_tree)
+
+    print('Done!')
+
+
+if __name__ == '__main__':
+
+    GTDB_tree_parser = argparse.ArgumentParser(usage=GTDB_tree_usage)
+    GTDB_tree_parser.add_argument('-p',               required=True,                            help='output prefix')
+    GTDB_tree_parser.add_argument('-i',               required=True,                            help='genome folder')
+    GTDB_tree_parser.add_argument('-x',               required=True,                            help='genome file extension')
+    GTDB_tree_parser.add_argument('-t',               required=False, type=int, default=1,      help='number of threads')
+    args = vars(GTDB_tree_parser.parse_args())
+    GTDB_tree(args)

TreeSAK/GeneTree.py

@@ -0,0 +1,142 @@
+import os
+import argparse
+from Bio import SeqIO
+from distutils.spawn import find_executable
+
+
+GeneTree_usage = '''
+============= GeneTree example commands =============
+
+TreeSAK GeneTree -i amoA.faa -o amoA_tree -t 36 -f
+
+=====================================================
+'''
+
+
+def check_dependencies(program_list):
+
+    not_detected_programs = []
+    for needed_program in program_list:
+        if find_executable(needed_program) is None:
+            not_detected_programs.append(needed_program)
+
+    if not_detected_programs != []:
+        print('%s not found, program exited!' % ','.join(not_detected_programs))
+        exit()
+
+
+def sep_path_basename_ext(file_in):
+
+    f_path, file_name = os.path.split(file_in)
+    if f_path == '':
+        f_path = '.'
+
+    f_base, f_ext = os.path.splitext(file_name)
+
+    return f_path, f_base, f_ext
+
+
+def filter_by_gap(file_in, max_gap_pct, file_out):
+    file_out_handle = open(file_out, 'w')
+    for each_seq in SeqIO.parse(file_in, 'fasta'):
+        seq_str = str(each_seq.seq)
+        gap_num = seq_str.count('-')
+        gap_pct = gap_num*100 / len(seq_str)
+        if gap_pct <= float(max_gap_pct):
+            file_out_handle.write('>%s\n%s\n' % (each_seq.id, seq_str))
+    file_out_handle.close()
+
+
+def GeneTree(args):
+
+    seq_file                    = args['i']
+    num_threads                 = args['t']
+    op_dir                      = args['o']
+    force_create_op_dir         = args['f']
+    gap_cutoff                  = args['max_gap']
+    trim_with_trimal            = args['trimal']
+    bmge_trim_model             = 'BLOSUM30'
+    bmge_entropy_score_cutoff   = '0.55'
+
+    # check dependencies
+    if trim_with_trimal is False:
+        check_dependencies(['mafft-einsi', 'java'])
+    else:
+        check_dependencies(['mafft-einsi', 'trimal'])
+
+    # specify path to BMGE.jar
+    current_file_path = '/'.join(os.path.realpath(__file__).split('/')[:-1])
+    pwd_bmge_jar      = '%s/BMGE.jar' % current_file_path
+
+    # determine the version of iqtree available on the system
+    if find_executable('iqtree2'):
+        iqtree_exe = 'iqtree2'
+    elif find_executable('iqtree'):
+        iqtree_exe = 'iqtree'
+    else:
+        print('iqtree not detected, program exited!')
+        exit()
+
+    # create op_dir
+    if os.path.isdir(op_dir) is True:
+        if force_create_op_dir is True:
+            os.system('rm -r %s' % op_dir)
+        else:
+            print('Output folder detected, program exited!')
+            exit()
+    os.system('mkdir %s' % op_dir)
+
+    ######################################## define output file name ########################################
+
+    sep_file_path, sep_file_base, sep_file_ext = sep_path_basename_ext(seq_file)
+    get_gene_tree_cmds_txt          = '%s/cmds.txt'                 % op_dir
+    msa_file                        = '%s/%s.aln'                   % (op_dir, sep_file_base)
+
+    msa_file_trimmed                = '%s/%s.bmge.aln'              % (op_dir, sep_file_base)
+    msa_file_trimmed_low_gap        = '%s/%s.bmge.maxgap%s.aln'     % (op_dir, sep_file_base, gap_cutoff)
+    if trim_with_trimal is True:
+        msa_file_trimmed            = '%s/%s.trimal.aln'            % (op_dir, sep_file_base)
+        msa_file_trimmed_low_gap    = '%s/%s.trimal.maxgap%s.aln'   % (op_dir, sep_file_base, gap_cutoff)
+
+    #########################################################################################################
+
+    # prepare commands
+    mafft_cmd       = 'mafft-einsi --thread %s --quiet %s > %s'              % (num_threads, seq_file, msa_file)
+    trim_cmd        = 'java -jar %s -i %s -m %s -t AA -h %s -of %s'          % (pwd_bmge_jar, msa_file, bmge_trim_model, bmge_entropy_score_cutoff, msa_file_trimmed)
+    if trim_with_trimal is True:
+        trim_cmd    = 'trimal -in %s -out %s -automated1'                    % (msa_file, msa_file_trimmed)
+    iqtree_cmd      = '%s -m LG+G+I -bb 1000 --wbtl -nt %s -s %s -pre %s/%s' % (iqtree_exe, num_threads, msa_file_trimmed_low_gap, op_dir, sep_file_base)
+
+    # write out commands
+    with open(get_gene_tree_cmds_txt, 'w') as f:
+        f.write('%s\n%s\n%s\n' % (mafft_cmd, trim_cmd, iqtree_cmd))
+
+    # run mafft
+    print(mafft_cmd)
+    os.system(mafft_cmd)
+
+    # run BMGE
+    print(trim_cmd)
+    os.system(trim_cmd)
+
+    # remove high gap sequences
+    filter_by_gap(msa_file_trimmed, gap_cutoff, msa_file_trimmed_low_gap)
+
+    # run iqtree
+    print(iqtree_cmd)
+    os.system(iqtree_cmd)
+
+    print('Done!')
+
+
+if __name__ == '__main__':
+
+    GeneTree_parser = argparse.ArgumentParser()
+    GeneTree_parser.add_argument('-i',          required=False, default=None,           help='sequence file')
+    GeneTree_parser.add_argument('-o',          required=True,                          help='output dir')
+    GeneTree_parser.add_argument('-t',          required=False, type=int, default=1,    help='number of threads, default is 1')
+    GeneTree_parser.add_argument('-trimal',     required=False, action="store_true",    help='trim with trimal, default is BMGE')
+    GeneTree_parser.add_argument('-f',          required=False, action="store_true",    help='force overwrite')
+    GeneTree_parser.add_argument('-max_gap',    required=False, default='40',           help='maximum percentage of gap, default is 40')
+    args = vars(GeneTree_parser.parse_args())
+    GeneTree(args)