def main(): info('correct-reads.py', ['streaming']) parser = get_parser() args = parser.parse_args() ### if len(set(args.input_filenames)) != len(args.input_filenames): print >>sys.stderr, \ "Error: Cannot input the same filename multiple times." sys.exit(1) ### report_on_config(args) check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savegraph: check_space_for_graph( args.n_tables * args.min_tablesize, args.force) K = args.ksize CUTOFF = args.cutoff NORMALIZE_LIMIT = args.normalize_to if args.loadgraph: print >>sys.stderr, 'loading k-mer countgraph from', args.loadgraph ct = khmer.load_countgraph(args.loadgraph) else: print >>sys.stderr, 'making k-mer countgraph' ct = khmer.new_countgraph(K, args.min_tablesize, args.n_tables) tempdir = tempfile.mkdtemp('khmer', 'tmp', args.tempdir) print >>sys.stderr, 'created temporary directory %s; ' \ 'use -T to change location' % tempdir aligner = khmer.ReadAligner(ct, args.cutoff, args.bits_theta) # ### FIRST PASS ### save_pass2_total = 0 n_bp = 0 n_reads = 0 written_bp = 0 written_reads = 0 corrected_reads = 0 pass2list = [] for filename in args.input_filenames: pass2filename = os.path.basename(filename) + '.pass2' pass2filename = os.path.join(tempdir, pass2filename) if args.out is None: corrfp = open(os.path.basename(filename) + '.corr', 'w') else: corrfp = args.out pass2list.append((filename, pass2filename, corrfp)) screed_iter = screed.open(filename, parse_description=False) pass2fp = open(pass2filename, 'w') save_pass2 = 0 n = 0 paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=args.ignore_pairs) for n, is_pair, read1, read2 in paired_iter: if n % 10000 == 0: print >>sys.stderr, '...', n, filename, save_pass2, \ n_reads, n_bp, written_reads, written_bp # we want to track paired reads here, to make sure that pairs # are not split between first pass and second pass. if is_pair: n_reads += 2 n_bp += len(read1.sequence) + len(read2.sequence) seq1 = read1.sequence.replace('N', 'A') seq2 = read2.sequence.replace('N', 'A') med1, _, _ = ct.get_median_count(seq1) med2, _, _ = ct.get_median_count(seq2) if med1 < NORMALIZE_LIMIT or med2 < NORMALIZE_LIMIT: ct.consume(seq1) ct.consume(seq2) write_record_pair(read1, read2, pass2fp) save_pass2 += 2 else: is_aligned, new_seq1 = correct_sequence(aligner, seq1) if is_aligned: if new_seq1 != read1.sequence: corrected_reads += 1 read1.sequence = new_seq1 if hasattr(read1, 'quality'): fix_quality(read1) is_aligned, new_seq2 = correct_sequence(aligner, seq2) if is_aligned: if new_seq2 != read2.sequence: corrected_reads += 1 read2.sequence = new_seq2 if hasattr(read2, 'quality'): fix_quality(read2) write_record_pair(read1, read2, corrfp) written_reads += 2 written_bp += len(read1) written_bp += len(read2) else: n_reads += 1 n_bp += len(read1.sequence) seq = read1.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # has this portion of the graph saturated? if not, # consume & save => pass2. if med < NORMALIZE_LIMIT: ct.consume(seq) write_record(read1, pass2fp) save_pass2 += 1 else: # trim!! is_aligned, new_seq = correct_sequence(aligner, seq) if is_aligned: if new_seq != read1.sequence: corrected_reads += 1 read1.sequence = new_seq if hasattr(read1, 'quality'): fix_quality(read1) write_record(read1, corrfp) written_reads += 1 written_bp += len(new_seq) pass2fp.close() print >>sys.stderr, '%s: kept aside %d of %d from first pass, in %s' \ % (filename, save_pass2, n, filename) save_pass2_total += save_pass2 # ### SECOND PASS. ### skipped_n = 0 skipped_bp = 0 for _, pass2filename, corrfp in pass2list: print >>sys.stderr, ('second pass: looking at sequences kept aside ' 'in %s') % pass2filename # note that for this second pass, we don't care about paired # reads - they will be output in the same order they're read in, # so pairs will stay together if not orphaned. This is in contrast # to the first loop. for n, read in enumerate(screed.open(pass2filename, parse_description=False)): if n % 10000 == 0: print >>sys.stderr, '... x 2', n, pass2filename, \ written_reads, written_bp seq = read.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # do we retain low-abundance components unchanged? if med < NORMALIZE_LIMIT and args.variable_coverage: write_record(read, corrfp) written_reads += 1 written_bp += len(read.sequence) skipped_n += 1 skipped_bp += len(read.sequence) # otherwise, examine/correct. else: # med >= NORMALIZE LIMIT or not args.variable_coverage is_aligned, new_seq = correct_sequence(aligner, seq) if is_aligned: if new_seq != read.sequence: corrected_reads += 1 read.sequence = new_seq if hasattr(read, 'quality'): fix_quality(read) write_record(read, corrfp) written_reads += 1 written_bp += len(new_seq) print >>sys.stderr, 'removing %s' % pass2filename os.unlink(pass2filename) print >>sys.stderr, 'removing temp directory & contents (%s)' % tempdir shutil.rmtree(tempdir) n_passes = 1.0 + (float(save_pass2_total) / n_reads) percent_reads_corrected = float(corrected_reads + (n_reads - written_reads)) /\ n_reads * 100.0 print >>sys.stderr, 'read %d reads, %d bp' % (n_reads, n_bp,) print >>sys.stderr, 'wrote %d reads, %d bp' % (written_reads, written_bp,) print >>sys.stderr, 'looked at %d reads twice (%.2f passes)' % \ (save_pass2_total, n_passes) print >>sys.stderr, 'removed %d reads and corrected %d reads (%.2f%%)' % \ (n_reads - written_reads, corrected_reads, percent_reads_corrected) print >>sys.stderr, 'removed %.2f%% of bases (%d total)' % \ ((1 - (written_bp / float(n_bp))) * 100.0, n_bp - written_bp) if args.variable_coverage: percent_reads_hicov = 100.0 * float(n_reads - skipped_n) / n_reads print >>sys.stderr, '%d reads were high coverage (%.2f%%);' % \ (n_reads - skipped_n, percent_reads_hicov) print >>sys.stderr, ('skipped %d reads/%d bases because of low' 'coverage') % (skipped_n, skipped_bp) fp_rate = \ khmer.calc_expected_collisions(ct, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print >>sys.stderr, \ 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) print >>sys.stderr, 'output in *.corr' if args.savegraph: print >>sys.stderr, "Saving k-mer countgraph to", args.savegraph ct.save(args.savegraph)
def main(): # pylint: disable=too-many-branches,too-many-statements info('normalize-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable(args.n_tables * args.min_tablesize, args.force) # list to save error files along with throwing exceptions if args.force: corrupt_files = [] if args.loadtable: print 'loading k-mer counting table from', args.loadtable htable = khmer.load_counting_hash(args.loadtable) else: print 'making k-mer counting table' htable = khmer.new_counting_hash(args.ksize, args.min_tablesize, args.n_tables) total = 0 discarded = 0 input_filename = None if args.single_output_filename: output_name = args.single_output_filename if args.append: outfp = open(args.single_output_filename, 'a') else: outfp = open(args.single_output_filename, 'w') for index, input_filename in enumerate(args.input_filenames): if not args.single_output_filename: output_name = os.path.basename(input_filename) + '.keep' outfp = open(output_name, 'w') total_acc = 0 discarded_acc = 0 try: total_acc, discarded_acc = normalize_by_median( input_filename, outfp, htable, args, report_fp) except IOError as err: handle_error(err, output_name, input_filename, args.fail_save, htable) if not args.force: print >> sys.stderr, '** Exiting!' sys.exit(1) else: print >> sys.stderr, '*** Skipping error file, moving on...' corrupt_files.append(input_filename) else: if total_acc == 0 and discarded_acc == 0: print 'SKIPPED empty file', input_filename else: total += total_acc discarded += discarded_acc print 'DONE with {inp}; kept {kept} of {total} or {perc:2}%'\ .format(inp=input_filename, kept=total - discarded, total=total, perc=int(100. - discarded / float(total) * 100.)) print 'output in', output_name if (args.dump_frequency > 0 and index > 0 and index % args.dump_frequency == 0): print 'Backup: Saving k-mer counting file through', input_filename if args.savetable: hashname = args.savetable print '...saving to', hashname else: hashname = 'backup.ct' print 'Nothing given for savetable, saving to', hashname htable.save(hashname) if args.report_total_kmers: print >> sys.stderr, 'Total number of unique k-mers: {0}'.format( htable.n_unique_kmers()) if args.savetable: print 'Saving k-mer counting table through', input_filename print '...saving to', args.savetable htable.save(args.savetable) fp_rate = khmer.calc_expected_collisions(htable) print 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) if args.force and len(corrupt_files) > 0: print >> sys.stderr, "** WARNING: Finished with errors!" print >> sys.stderr, "** IOErrors occurred in the following files:" print >> sys.stderr, "\t", " ".join(corrupt_files) if fp_rate > MAX_FALSE_POSITIVE_RATE: print >> sys.stderr, "**" print >> sys.stderr, ("** ERROR: the k-mer counting table is too small" " for this data set. Increase tablesize/# " "tables.") print >> sys.stderr, "**" print >> sys.stderr, "** Do not use these results!!" if not args.force: sys.exit(1)
def main(): # pylint: disable=too-many-branches,too-many-statements start_time = time.time() parser = sanitize_help(get_parser()) args = parser.parse_args() configure_logging(args.quiet) report_on_config(args) report_fp = args.report force_single = args.force_single # check for similar filenames # if we're using a single output file only check for identical filenames # otherwise, check for identical BASE names as well. filenames = [] basenames = [] for pathfilename in args.input_filenames: filenames.append(pathfilename) if args.single_output_file: continue # nothing more to worry about basename = os.path.basename(pathfilename) if basename in basenames: log_error('ERROR: Duplicate filename--Cannot handle this!') log_error('** Exiting!') sys.exit(1) basenames.append(basename) # check that files exist and there is sufficient output disk space. check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savegraph is not None: graphsize = calculate_graphsize(args, 'countgraph') check_space_for_graph(args.savegraph, graphsize, args.force) # load or create counting table. if args.loadgraph: log_info('loading k-mer countgraph from {graph}', graph=args.loadgraph) countgraph1 = Countgraph.load(args.loadgraph) # load second counting table. if args.loadgraph2: log_info('loading k-mer countgraph from {graph}', graph=args.loadgraph2) countgraph2 = Countgraph.load(args.loadgraph2) # make a list of all filenames and if they're paired or not; # if we don't know if they're paired, default to allowing but not # forcing pairing. files = [] for element in filenames: files.append([element, args.paired]) if args.unpaired_reads: files.append([args.unpaired_reads, False]) # # main loop: iterate over all files given, do diginorm. # for filename, require_paired in files: if not args.single_output_file: output_name = os.path.basename(filename) + '.keep' outfp = open(output_name, 'wb') outfp = get_file_writer(outfp, args.gzip, args.bzip) screed_iter = clean_input_reads(screed.open(filename)) reader = broken_paired_reader(screed_iter, min_length=args.ksize, force_single=force_single, require_paired=require_paired) # actually do diginorm for _, is_paired, read0, read1 in reader: for record in snarf(is_paired, read0, read1, countgraph1, countgraph2): if record is not None: write_record(record, outfp) print("--- %s seconds ---" % (time.time() - start_time))
def main(): # pylint: disable=too-many-branches,too-many-statements info('normalize-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report # check for similar filenames filenames = [] for pathfilename in args.input_filenames: filename = pathfilename.split('/')[-1] if (filename in filenames): print >>sys.stderr, "WARNING: At least two input files are named \ %s . (The script normalize-by-median.py can not handle this, only one .keep \ file for one of the input files will be generated.)" % filename else: filenames.append(filename) # check for others check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable( args.n_tables * args.min_tablesize, args.force) # list to save error files along with throwing exceptions if args.force: corrupt_files = [] if args.loadtable: print 'loading k-mer counting table from', args.loadtable htable = khmer.load_counting_hash(args.loadtable) else: print >> sys.stderr, 'making k-mer counting table' htable = khmer.new_counting_hash(args.ksize, args.min_tablesize, args.n_tables) total = 0 discarded = 0 input_filename = None if args.single_output_file: outfp = args.single_output_file if args.single_output_file is sys.stdout: output_name = '/dev/stdout' else: output_name = args.single_output_file.name for index, input_filename in enumerate(args.input_filenames): if not args.single_output_file: output_name = os.path.basename(input_filename) + '.keep' outfp = open(output_name, 'w') total_acc = 0 discarded_acc = 0 try: total_acc, discarded_acc = normalize_by_median(input_filename, outfp, htable, args, report_fp) except IOError as err: handle_error(err, output_name, input_filename, args.fail_save, htable) if not args.force: print >> sys.stderr, '** Exiting!' sys.exit(1) else: print >> sys.stderr, '*** Skipping error file, moving on...' corrupt_files.append(input_filename) else: if total_acc == 0 and discarded_acc == 0: print >> sys.stderr, 'SKIPPED empty file', input_filename else: total += total_acc discarded += discarded_acc print >> sys.stderr, \ 'DONE with {inp}; kept {kept} of {total} or {perc:2}%'\ .format(inp=input_filename, kept=total - discarded, total=total, perc=int(100. - discarded / float(total) * 100.)) print >> sys.stderr, 'output in', output_name if (args.dump_frequency > 0 and index > 0 and index % args.dump_frequency == 0): print 'Backup: Saving k-mer counting file through', input_filename if args.savetable: hashname = args.savetable print '...saving to', hashname else: hashname = 'backup.ct' print 'Nothing given for savetable, saving to', hashname htable.save(hashname) if args.report_total_kmers: print >> sys.stderr, 'Total number of unique k-mers: {0}'.format( htable.n_unique_kmers()) if args.savetable: print 'Saving k-mer counting table through', input_filename print '...saving to', args.savetable htable.save(args.savetable) fp_rate = \ khmer.calc_expected_collisions(htable, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print >> sys.stderr, \ 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) if args.force and len(corrupt_files) > 0: print >> sys.stderr, "** WARNING: Finished with errors!" print >> sys.stderr, "** IOErrors occurred in the following files:" print >> sys.stderr, "\t", " ".join(corrupt_files)
def main(): parser = argparse.ArgumentParser() parser.add_argument('database') parser.add_argument('input_filenames', metavar='input_sequence_filename', help='Input FAST[AQ] sequence filename.', nargs='+') parser.add_argument('-k', '--ksize', type=int, default=31) parser.add_argument('-p', '--paired', action='store_true', help='require that all sequences be properly paired') parser.add_argument('--force_single', dest='force_single', action='store_true', help='treat all sequences as single-ended/unpaired') parser.add_argument('-u', '--unpaired-reads', metavar="unpaired_reads_filename", help='include a file of unpaired reads to which ' '-p/--paired does not apply.') parser.add_argument('-f', '--force', dest='force', help='continue past file reading errors', action='store_true') args = parser.parse_args() force_single = args.force_single #if args.reads == '-': # args.reads = sys.stdin # check that input files exist check_valid_file_exists(args.input_filenames) filenames = [] for pathfilename in args.input_filenames: filenames.append(pathfilename) # make a list of all filenames and if they're paired or not; # if we don't know if they're paired, default to allowing but not # forcing pairing. files = [] for element in filenames: files.append([element, args.paired]) if args.unpaired_reads: files.append([args.unpaired_reads, False]) # create object of Nodetable in Khmer to use its kh = khmer.Nodetable(args.ksize, 1, 1) # load database mphf_filename = args.database + '.mphf' array_filename = args.database + '.arr' print('loading database {}'.format(args.database)) with open(array_filename, 'rb') as fp: mphf_to_kmer, mphf_to_cdbg, family_ids, cdbg_to_family_id = pickle.load( fp) mphf = bbhash.load_mphf(mphf_filename) print('done!') def get_kmer_to_family_ids(hashval): mphf_hash = mphf.lookup(hashval) if mphf_hash is None: return set() kmer_hash = mphf_to_kmer[mphf_hash] if kmer_hash != hashval: return set() cdbg_id = mphf_to_cdbg[mphf_hash] id_list = cdbg_to_family_id[cdbg_id] return id_list def readFusion(read): global n_unmatched, n_same, n_amb_same, n_clear_fusion, n_ambig_fusion, n_mutli_fusion flag = None lf_ids = set() rt_ids = set() families = [] shared_kmers = [] gaps = [] hashvals = kh.get_kmer_hashes(read.sequence) # find a matching k-mer at the beginning of the read lf = hashvals[0] lf_ids = get_kmer_to_family_ids(lf) idx = 1 while idx < len(hashvals) and len(lf_ids) == 0: lf = hashvals[idx] lf_ids = get_kmer_to_family_ids(lf) idx += 1 if len(lf_ids) == 0: #print('no single match') n_unmatched += 1 flag = "unmatched" elif idx == len(hashvals): #print('same, only last kmer matched') families.append(lf_ids) if len(lf_ids) == 1: n_same += 1 flag = "unique" else: n_amb_same += 1 flag = "ambiguous" else: # len(lf_ids) > 0 & idx < len(hashvals) # find a matching k-mer at the end of the read rt = hashvals[-1] rt_ids = get_kmer_to_family_ids(rt) idy = len(hashvals) - 2 while idy >= idx and len(rt_ids) == 0: rt = hashvals[idy] rt_ids = get_kmer_to_family_ids(rt) idy -= 1 if len(rt_ids) == 0: #print('same, only one non-last kmer matched ') families.append(lf_ids) if len(lf_ids) == 1: n_same += 1 flag = "unique" else: n_amb_same += 1 flag = "ambiguous" else: intersect_ids = lf_ids.intersection(rt_ids) if len(intersect_ids) > 0: families.append(intersect_ids) if len(intersect_ids) == 1: n_same += 1 flag = "unique" else: n_amb_same += 1 flag = "ambiguous" else: # fusion to be resolved shared_kmer = 1 gap_size = 0 gap = False while idx <= idy + 1: temp = hashvals[idx] temp_ids = get_kmer_to_family_ids(temp) if len(temp_ids) > 0: intersect_ids = lf_ids.intersection(temp_ids) if len(intersect_ids) > 0: lf_ids = intersect_ids shared_kmer += 1 gap_size = 0 else: # len(intersect_ids) == 0 families.append(lf_ids) shared_kmers.append(shared_kmer) lf_ids = temp_ids shared_kmer = 1 gaps.append(gap_size) gap_size = 0 else: gap_size += 1 idx += 1 families.append(lf_ids) shared_kmers.append(shared_kmer) assert len(families) > 1 if len(families) == 2: if len(families[0]) == 1 and len(families[1]) == 1: n_clear_fusion += 1 flag = "clear_fusion" else: n_ambig_fusion += 1 flag = "ambig_fusion" else: # len(families) > 2 n_mutli_fusion += 1 flag = "multi_fusion" #if len(families) == 0: # families = "-" #if len(shared_kmers) == 0: # shared_kmers = "-" return flag, families, shared_kmers, gaps fusion_filename = args.database + '_fusion.fa' fusion_fp = open(fusion_filename, 'w') fusionInfo_filename = args.database + '_fusion.info' fusionInfo_fp = open(fusionInfo_filename, 'w') print("fileName", "recordIndex", "whichInPair", "align_class", "gene_families", "shared_kmers", "gaps", file=fusionInfo_fp, sep='\t') fusionCalc_filename = args.database + '_fusion.calc' fusionCalc_fp = open(fusionCalc_filename, 'w') print("fileName", "recordIndex", "whichInPair", "align_class", "familiy_A", "familiy_B", "no_families", "len_families", "shared_kmers", "gaps", "sorted_keys", file=fusionCalc_fp, sep='\t') fusionPairs_filename = args.database + '_fusionPairs.fa' fusPair_fp = open(fusionPairs_filename, 'w') fusionPairsInfo_filename = args.database + '_fusionPairs.info' fusPairInfo_fp = open(fusionPairsInfo_filename, 'w') print("fileName", "recordIndex", "fusion_class", "R1_family", "R2_family", file=fusPairInfo_fp, sep='\t') fusionPairsCalc_filename = args.database + '_fusionPairs.calc' fusPairCalc_fp = open(fusionPairsCalc_filename, 'w') print("fileName", "recordIndex", "fusion_class", "familiy_A", "familiy_B", "len_families", "sorted_keys", file=fusPairCalc_fp, sep='\t') corrupt_files = [] family_names = dict(zip(family_ids.values(), family_ids.keys())) n = 0 n_paired_fusion = 0 sameRef = ("unique", "ambiguous") fusion = ("clear_fusion", "ambig_fusion", "multi_fusion") for filename, require_paired in files: with catch_io_errors(filename, fusion_fp, fusionInfo_fp, fusionCalc_fp, fusPair_fp, fusPairInfo_fp, fusPairCalc_fp, args.force, corrupt_files): screed_iter = clean_input_reads(screed.open(filename)) reader = broken_paired_reader(screed_iter, min_length=args.ksize, force_single=force_single, require_paired=require_paired) for r_index, is_paired, read0, read1 in reader: n += 1 if n % 10000 == 0: print('...', n) #if n > 5000: # break flag0, families0, shared_kmers0, gaps0 = readFusion(read0) if not is_paired and flag0 in fusion: #families_names0 = [] #for gp in families0: # gp_names = [] # for family_id in gp: # family_name = family_names[family_id] # gp_names.append(family_name) # families_names0.append(gp_names) print(filename, r_index, "single", flag0, families0, shared_kmers0, gaps0, file=fusionInfo_fp, sep='\t') write_record(read0, fusion_fp) #i = 1 #while i < len(families0): # for g1 in families0[i-1]: # for g2 in families0[i]: # print(filename, r_index, "single", flag0, sorted([g1,g2]), len(families0), len(families0[i-1]), len(families0[i]), # shared_kmers0, gaps0, file=fusionCalc_fp, sep='\t') # i += 1 i = len(families0) - 1 for g1 in families0[0]: g1_name = family_names[g1] for g2 in families0[i]: g2_name = family_names[g2] print(filename, r_index, "single", flag0, '{}:{}'.format(g1, g1_name), '{}:{}'.format(g2, g2_name), len(families0), [len(f) for f in families0], shared_kmers0, gaps0, sorted([g1, g2]), file=fusionCalc_fp, sep='\t') if is_paired: flag1, families1, shared_kmers1, gaps1 = readFusion(read1) if flag0 in fusion or flag1 in fusion: print(filename, r_index, "Read_1", flag0, families0, shared_kmers0, gaps0, file=fusionInfo_fp, sep='\t') write_record(read0, fusion_fp) print(filename, r_index, "Read_2", flag1, families1, shared_kmers1, gaps1, file=fusionInfo_fp, sep='\t') write_record(read1, fusion_fp) if flag0 in fusion: i = len(families0) - 1 for g1 in families0[0]: g1_name = family_names[g1] for g2 in families0[i]: g2_name = family_names[g2] print(filename, r_index, "Read_1", flag0, '{}:{}'.format(g1, g1_name), '{}:{}'.format(g2, g2_name), len(families0), [len(f) for f in families0], shared_kmers0, gaps0, sorted([g1, g2]), file=fusionCalc_fp, sep='\t') if flag1 in fusion: i = len(families1) - 1 for g1 in families1[0]: g1_name = family_names[g1] for g2 in families1[i]: g2_name = family_names[g2] print(filename, r_index, "Read_2", flag1, '{}:{}'.format(g1, g1_name), '{}:{}'.format(g2, g2_name), len(families1), [len(f) for f in families1], shared_kmers1, gaps1, sorted([g1, g2]), file=fusionCalc_fp, sep='\t') elif flag0 in sameRef and flag1 in sameRef: if len(families0[0].intersection(families1[0])) == 0: n_paired_fusion += 1 if flag0 == "unique" and flag1 == "unique": fusion_class = "clear_fusion" else: fusion_class = "ambig_fusion" print(filename, r_index, fusion_class, families0, families1, file=fusPairInfo_fp, sep='\t') write_record(read0, fusPair_fp) write_record(read1, fusPair_fp) for g1 in families0[0]: g1_name = family_names[g1] for g2 in families1[0]: g2_name = family_names[g2] print(filename, r_index, fusion_class, '{}:{}'.format(g1, g1_name), '{}:{}'.format(g2, g2_name), [ len(f) for f in (families0[0], families1[0]) ], sorted([g1, g2]), file=fusPairCalc_fp, sep='\t') print('No of input fragments: ', n) print('unmatched:', n_unmatched) print('Unique:', n_same) print('Ambiguous:', n_amb_same) print('Single read clear fusion:', n_clear_fusion) print('Single read ambiguous fusion:', n_ambig_fusion) print('Single read multi fusion:', n_mutli_fusion) print('paired read fusion:', n_paired_fusion)
def main(): # pylint: disable=too-many-branches,too-many-statements parser = sanitize_help(get_parser()) args = parser.parse_args() configure_logging(args.quiet) report_on_config(args) report_fp = args.report force_single = args.force_single # check for similar filenames # if we're using a single output file only check for identical filenames # otherwise, check for identical BASE names as well. filenames = [] basenames = [] for pathfilename in args.input_filenames: filenames.append(pathfilename) if args.single_output_file: continue # nothing more to worry about basename = os.path.basename(pathfilename) if basename in basenames: log_error('ERROR: Duplicate filename--Cannot handle this!') log_error('** Exiting!') sys.exit(1) basenames.append(basename) # check that files exist and there is sufficient output disk space. check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savegraph is not None: graphsize = calculate_graphsize(args, 'countgraph') check_space_for_graph(args.savegraph, graphsize, args.force) # load or create counting table. if args.loadgraph: log_info('loading k-mer countgraph from {graph}', graph=args.loadgraph) countgraph = khmer.load_countgraph(args.loadgraph) else: log_info('making countgraph') countgraph = khmer_args.create_countgraph(args) # create an object to handle diginorm of all files norm = Normalizer(args.cutoff, countgraph) with_diagnostics = WithDiagnostics(norm, report_fp, args.report_frequency) # make a list of all filenames and if they're paired or not; # if we don't know if they're paired, default to allowing but not # forcing pairing. files = [] for element in filenames: files.append([element, args.paired]) if args.unpaired_reads: files.append([args.unpaired_reads, False]) corrupt_files = [] outfp = None output_name = None if args.single_output_file: outfp = get_file_writer(args.single_output_file, args.gzip, args.bzip) else: if '-' in filenames or '/dev/stdin' in filenames: print( "Accepting input from stdin; output filename must " "be provided with '-o'.", file=sys.stderr) sys.exit(1) # # main loop: iterate over all files given, do diginorm. # for filename, require_paired in files: if not args.single_output_file: output_name = os.path.basename(filename) + '.keep' outfp = open(output_name, 'wb') outfp = get_file_writer(outfp, args.gzip, args.bzip) # failsafe context manager in case an input file breaks with catch_io_errors(filename, outfp, args.single_output_file, args.force, corrupt_files): screed_iter = clean_input_reads(screed.open(filename)) reader = broken_paired_reader(screed_iter, min_length=args.ksize, force_single=force_single, require_paired=require_paired) # actually do diginorm for record in with_diagnostics(reader, filename): if record is not None: write_record(record, outfp) log_info('output in {name}', name=describe_file_handle(outfp)) if not args.single_output_file: outfp.close() # finished - print out some diagnostics. log_info('Total number of unique k-mers: {umers}', umers=countgraph.n_unique_kmers()) if args.savegraph is not None: log_info('...saving to {name}', name=args.savegraph) countgraph.save(args.savegraph) fp_rate = \ khmer.calc_expected_collisions(countgraph, False, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 log_info('fp rate estimated to be {fpr:1.3f}', fpr=fp_rate) if args.force and len(corrupt_files) > 0: log_error("** WARNING: Finished with errors!") log_error("** I/O Errors occurred in the following files:") log_error("\t" + " ".join(corrupt_files))
def main(): # pylint: disable=too-many-branches,too-many-statements info('normalize-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report force_single = args.force_single # check for similar filenames # if we're using a single output file only check for identical filenames # otherwise, check for identical BASE names as well. filenames = [] basenames = [] for pathfilename in args.input_filenames: filenames.append(pathfilename) if args.single_output_file: continue # nothing more to worry about basename = os.path.basename(pathfilename) if basename in basenames: print('ERROR: Duplicate filename--Cannot handle this!', file=sys.stderr) print('** Exiting!', file=sys.stderr) sys.exit(1) basenames.append(basename) # check that files exist and there is sufficient output disk space. check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable(args, 'countgraph', args.force) # load or create counting table. if args.loadtable: print('loading k-mer counting table from ' + args.loadtable, file=sys.stderr) htable = khmer.load_counting_hash(args.loadtable) else: print('making countgraph', file=sys.stderr) htable = khmer_args.create_countgraph(args) input_filename = None # create an object to handle diginorm of all files norm = Normalizer(args.cutoff, htable) # make a list of all filenames and if they're paired or not; # if we don't know if they're paired, default to allowing but not # forcing pairing. files = [] for e in filenames: files.append([e, args.paired]) if args.unpaired_reads: files.append([args.unpaired_reads, False]) corrupt_files = [] outfp = None output_name = None if args.single_output_file: if args.single_output_file is sys.stdout: output_name = '/dev/stdout' else: output_name = args.single_output_file.name outfp = args.single_output_file # # main loop: iterate over all files given, do diginorm. # for filename, require_paired in files: if not args.single_output_file: output_name = os.path.basename(filename) + '.keep' outfp = open(output_name, 'w') # failsafe context manager in case an input file breaks with CatchIOErrors(filename, outfp, args.single_output_file, args.force, corrupt_files): screed_iter = screed.open(filename, parse_description=False) reader = broken_paired_reader(screed_iter, min_length=args.ksize, force_single=force_single, require_paired=require_paired) # actually do diginorm for record in WithDiagnostics(filename, norm, reader, report_fp): if record is not None: write_record(record, outfp) print('output in ' + output_name, file=sys.stderr) if output_name is not '/dev/stdout': outfp.close() # finished - print out some diagnostics. print('Total number of unique k-mers: {0}' .format(htable.n_unique_kmers()), file=sys.stderr) if args.savetable: print('...saving to ' + args.savetable, file=sys.stderr) htable.save(args.savetable) fp_rate = \ khmer.calc_expected_collisions(htable, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print('fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate), file=sys.stderr) if args.force and len(corrupt_files) > 0: print("** WARNING: Finished with errors!", file=sys.stderr) print("** I/O Errors occurred in the following files:", file=sys.stderr) print("\t", " ".join(corrupt_files), file=sys.stderr)
def main(): # pylint: disable=too-many-branches,too-many-statements parser = sanitize_help(get_parser()) args = parser.parse_args() configure_logging(args.quiet) report_on_config(args) report_fp = args.report force_single = args.force_single # check for similar filenames # if we're using a single output file only check for identical filenames # otherwise, check for identical BASE names as well. filenames = [] basenames = [] for pathfilename in args.input_filenames: filenames.append(pathfilename) if args.single_output_file: continue # nothing more to worry about basename = os.path.basename(pathfilename) if basename in basenames: log_error('ERROR: Duplicate filename--Cannot handle this!') log_error('** Exiting!') sys.exit(1) basenames.append(basename) # check that files exist and there is sufficient output disk space. check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savegraph is not None: graphsize = calculate_graphsize(args, 'countgraph') check_space_for_graph(args.savegraph, graphsize, args.force) # load or create counting table. if args.loadgraph: log_info('loading k-mer countgraph from {graph}', graph=args.loadgraph) countgraph = Countgraph.load(args.loadgraph) else: log_info('making countgraph') countgraph = khmer_args.create_countgraph(args) # create an object to handle diginorm of all files norm = Normalizer(args.cutoff, countgraph) with_diagnostics = WithDiagnostics(norm, report_fp, args.report_frequency) # make a list of all filenames and if they're paired or not; # if we don't know if they're paired, default to allowing but not # forcing pairing. files = [] for element in filenames: files.append([element, args.paired]) if args.unpaired_reads: files.append([args.unpaired_reads, False]) corrupt_files = [] outfp = None output_name = None if args.single_output_file: outfp = get_file_writer(args.single_output_file, args.gzip, args.bzip) else: if '-' in filenames or '/dev/stdin' in filenames: print("Accepting input from stdin; output filename must " "be provided with '-o'.", file=sys.stderr) sys.exit(1) # # main loop: iterate over all files given, do diginorm. # for filename, require_paired in files: if not args.single_output_file: output_name = os.path.basename(filename) + '.keep' outfp = open(output_name, 'wb') outfp = get_file_writer(outfp, args.gzip, args.bzip) # failsafe context manager in case an input file breaks with catch_io_errors(filename, outfp, args.single_output_file, args.force, corrupt_files): screed_iter = clean_input_reads(screed.open(filename)) reader = broken_paired_reader(screed_iter, min_length=args.ksize, force_single=force_single, require_paired=require_paired) # actually do diginorm for record in with_diagnostics(reader, filename): if record is not None: write_record(record, outfp) log_info('output in {name}', name=describe_file_handle(outfp)) if not args.single_output_file: outfp.close() # finished - print out some diagnostics. log_info('Total number of unique k-mers: {umers}', umers=countgraph.n_unique_kmers()) if args.savegraph is not None: log_info('...saving to {name}', name=args.savegraph) countgraph.save(args.savegraph) fp_rate = \ khmer.calc_expected_collisions(countgraph, False, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 log_info('fp rate estimated to be {fpr:1.3f}', fpr=fp_rate) if args.force and len(corrupt_files) > 0: log_error("** WARNING: Finished with errors!") log_error("** I/O Errors occurred in the following files:") log_error("\t" + " ".join(corrupt_files))
def main(): parser = sanitize_help(get_parser()) args = parser.parse_args() if not args.quiet: info('trim-low-abund.py', ['streaming']) configure_logging(args.quiet) ### if len(set(args.input_filenames)) != len(args.input_filenames): log_error("Error: Cannot input the same filename multiple times.") sys.exit(1) if args.trim_at_coverage != DEFAULT_TRIM_AT_COVERAGE and \ not args.variable_coverage: log_error("Error: --trim-at-coverage/-Z given, but " "--variable-coverage/-V not specified.") sys.exit(1) if args.diginorm_coverage != DEFAULT_DIGINORM_COVERAGE and \ not args.diginorm: log_error("Error: --diginorm-coverage given, but " "--diginorm not specified.") sys.exit(1) if args.diginorm and args.single_pass: log_error("Error: --diginorm and --single-pass are incompatible!\n" "You probably want to use normalize-by-median.py instead.") sys.exit(1) ### report_on_config(args) check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savegraph: graphsize = calculate_graphsize(args, 'countgraph') check_space_for_graph(args.savegraph, graphsize, args.force) if ('-' in args.input_filenames or '/dev/stdin' in args.input_filenames) \ and not args.output: log_error("Accepting input from stdin; output filename must " "be provided with -o.") sys.exit(1) if args.loadgraph: log_info('loading countgraph from {graph}', graph=args.loadgraph) ct = khmer.load_countgraph(args.loadgraph) else: log_info('making countgraph') ct = khmer_args.create_countgraph(args) K = ct.ksize() tempdir = tempfile.mkdtemp('khmer', 'tmp', args.tempdir) log_info('created temporary directory {temp};\n' 'use -T to change location', temp=tempdir) trimmer = Trimmer(ct, not args.variable_coverage, args.cutoff, args.trim_at_coverage) if args.diginorm: trimmer.set_diginorm(args.diginorm_coverage) # ### FIRST PASS ### save_pass2_total = 0 written_bp = 0 written_reads = 0 # only create the file writer once if outfp is specified; otherwise, # create it for each file. if args.output: trimfp = get_file_writer(args.output, args.gzip, args.bzip) pass2list = [] for filename in args.input_filenames: # figure out temporary filename for 2nd pass pass2filename = os.path.basename(filename) + '.pass2' pass2filename = os.path.join(tempdir, pass2filename) pass2fp = open(pass2filename, 'w') # construct output filenames if args.output is None: # note: this will be saved in trimfp. outfp = open(os.path.basename(filename) + '.abundtrim', 'wb') # get file handle w/gzip, bzip trimfp = get_file_writer(outfp, args.gzip, args.bzip) # record all this info pass2list.append((filename, pass2filename, trimfp)) # input file stuff: get a broken_paired reader. screed_iter = screed.open(filename) paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=args.ignore_pairs) # main loop through the file. n_start = trimmer.n_reads save_start = trimmer.n_saved watermark = REPORT_EVERY_N_READS for read in trimmer.pass1(paired_iter, pass2fp): if (trimmer.n_reads - n_start) > watermark: log_info("... {filename} {n_saved} {n_reads} {n_bp} " "{w_reads} {w_bp}", filename=filename, n_saved=trimmer.n_saved, n_reads=trimmer.n_reads, n_bp=trimmer.n_bp, w_reads=written_reads, w_bp=written_bp) watermark += REPORT_EVERY_N_READS # write out the trimmed/etc sequences that AREN'T going to be # revisited in a 2nd pass. write_record(read, trimfp) written_bp += len(read) written_reads += 1 pass2fp.close() log_info("{filename}: kept aside {kept} of {total} from first pass", filename=filename, kept=trimmer.n_saved - save_start, total=trimmer.n_reads - n_start) # first pass goes across all the data, so record relevant stats... n_reads = trimmer.n_reads n_bp = trimmer.n_bp n_skipped = trimmer.n_skipped bp_skipped = trimmer.bp_skipped save_pass2_total = trimmer.n_saved # ### SECOND PASS. ### # nothing should have been skipped yet! assert trimmer.n_skipped == 0 assert trimmer.bp_skipped == 0 if args.single_pass: pass2list = [] # go back through all the files again. for _, pass2filename, trimfp in pass2list: log_info('second pass: looking at sequences kept aside in {pass2}', pass2=pass2filename) # note that for this second pass, we don't care about paired # reads - they will be output in the same order they're read in, # so pairs will stay together if not orphaned. This is in contrast # to the first loop. Hence, force_single=True below. screed_iter = screed.open(pass2filename, parse_description=False) paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=True) watermark = REPORT_EVERY_N_READS for read in trimmer.pass2(paired_iter): if (trimmer.n_reads - n_start) > watermark: log_info('... x 2 {a} {b} {c} {d} {e} {f} {g}', a=trimmer.n_reads - n_start, b=pass2filename, c=trimmer.n_saved, d=trimmer.n_reads, e=trimmer.n_bp, f=written_reads, g=written_bp) watermark += REPORT_EVERY_N_READS write_record(read, trimfp) written_reads += 1 written_bp += len(read) log_info('removing {pass2}', pass2=pass2filename) os.unlink(pass2filename) # if we created our own trimfps, close 'em. if not args.output: trimfp.close() log_info('removing temp directory & contents ({temp})', temp=tempdir) shutil.rmtree(tempdir) trimmed_reads = trimmer.trimmed_reads n_passes = 1.0 + (float(save_pass2_total) / n_reads) percent_reads_trimmed = float(trimmed_reads + (n_reads - written_reads)) /\ n_reads * 100.0 log_info('read {read} reads, {bp} bp', read=n_reads, bp=n_bp) log_info('wrote {wr} reads, {wbp} bp', wr=written_reads, wbp=written_bp) log_info('looked at {st} reads twice ({np:.2f} passes)', st=save_pass2_total, np=n_passes) log_info('removed {r} reads and trimmed {t} reads ({p:.2f}%)', r=n_reads - written_reads, t=trimmed_reads, p=percent_reads_trimmed) log_info('trimmed or removed {p:.2f}%% of bases ({bp} total)', p=(1 - (written_bp / float(n_bp))) * 100.0, bp=n_bp - written_bp) if args.variable_coverage: percent_reads_hicov = 100.0 * float(n_reads - n_skipped) / n_reads log_info('{n} reads were high coverage ({p:.2f}%);', n=n_reads - n_skipped, p=percent_reads_hicov) log_info('skipped {r} reads/{bp} bases because of low coverage', r=n_skipped, bp=bp_skipped) fp_rate = \ khmer.calc_expected_collisions(ct, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 log_info('fp rate estimated to be {fpr:1.3f}', fpr=fp_rate) log_info('output in *.abundtrim') if args.savegraph: log_info("Saving k-mer countgraph to {graph}", graph=args.savegraph) ct.save(args.savegraph)
def main(): info('correct-reads.py', ['streaming']) args = sanitize_help(get_parser()).parse_args() ### if len(set(args.input_filenames)) != len(args.input_filenames): print("Error: Cannot input the same filename multiple times.", file=sys.stderr) sys.exit(1) ### report_on_config(args) check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) tablesize = calculate_graphsize(args, 'countgraph') if args.savegraph: check_space_for_graph(args.savegraph, tablesize, args.force) K = args.ksize CUTOFF = args.cutoff NORMALIZE_LIMIT = args.normalize_to if args.loadgraph: print('loading k-mer countgraph from', args.loadgraph, file=sys.stderr) ct = Countgraph.load(args.loadgraph) else: print('making k-mer countgraph', file=sys.stderr) ct = create_countgraph(args, multiplier=8 / (9. + 0.3)) tempdir = tempfile.mkdtemp('khmer', 'tmp', args.tempdir) print('created temporary directory %s; use -T to change location' % tempdir, file=sys.stderr) aligner = khmer.ReadAligner(ct, args.cutoff, args.bits_theta) # ### FIRST PASS ### save_pass2_total = 0 n_bp = 0 n_reads = 0 written_bp = 0 written_reads = 0 corrected_reads = 0 pass2list = [] for filename in args.input_filenames: pass2filename = os.path.basename(filename) + '.pass2' pass2filename = os.path.join(tempdir, pass2filename) if args.out is None: corrfp = open(os.path.basename(filename) + '.corr', 'w') else: corrfp = args.out pass2list.append((filename, pass2filename, corrfp)) screed_iter = screed.open(filename, parse_description=False) pass2fp = open(pass2filename, 'w') save_pass2 = 0 n = 0 paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=args.ignore_pairs) for n, is_pair, read1, read2 in paired_iter: if n % 10000 == 0: print('...', n, filename, save_pass2, n_reads, n_bp, written_reads, written_bp, file=sys.stderr) # we want to track paired reads here, to make sure that pairs # are not split between first pass and second pass. if is_pair: n_reads += 2 n_bp += len(read1.sequence) + len(read2.sequence) seq1 = read1.sequence.replace('N', 'A') seq2 = read2.sequence.replace('N', 'A') med1, _, _ = ct.get_median_count(seq1) med2, _, _ = ct.get_median_count(seq2) if med1 < NORMALIZE_LIMIT or med2 < NORMALIZE_LIMIT: ct.consume(seq1) ct.consume(seq2) write_record_pair(read1, read2, pass2fp) save_pass2 += 2 else: is_aligned, new_seq1 = correct_sequence(aligner, seq1) if is_aligned: if new_seq1 != read1.sequence: corrected_reads += 1 read1.sequence = new_seq1 if hasattr(read1, 'quality'): fix_quality(read1) is_aligned, new_seq2 = correct_sequence(aligner, seq2) if is_aligned: if new_seq2 != read2.sequence: corrected_reads += 1 read2.sequence = new_seq2 if hasattr(read2, 'quality'): fix_quality(read2) write_record_pair(read1, read2, corrfp) written_reads += 2 written_bp += len(read1) written_bp += len(read2) else: n_reads += 1 n_bp += len(read1.sequence) seq = read1.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # has this portion of the graph saturated? if not, # consume & save => pass2. if med < NORMALIZE_LIMIT: ct.consume(seq) write_record(read1, pass2fp) save_pass2 += 1 else: # trim!! is_aligned, new_seq = correct_sequence(aligner, seq) if is_aligned: if new_seq != read1.sequence: corrected_reads += 1 read1.sequence = new_seq if hasattr(read1, 'quality'): fix_quality(read1) write_record(read1, corrfp) written_reads += 1 written_bp += len(new_seq) pass2fp.close() print('%s: kept aside %d of %d from first pass, in %s' % (filename, save_pass2, n, filename), file=sys.stderr) save_pass2_total += save_pass2 # ### SECOND PASS. ### skipped_n = 0 skipped_bp = 0 for _, pass2filename, corrfp in pass2list: print(('second pass: looking at sequences kept aside in %s') % pass2filename, file=sys.stderr) # note that for this second pass, we don't care about paired # reads - they will be output in the same order they're read in, # so pairs will stay together if not orphaned. This is in contrast # to the first loop. for n, read in enumerate( screed.open(pass2filename, parse_description=False)): if n % 10000 == 0: print('... x 2', n, pass2filename, written_reads, written_bp, file=sys.stderr) seq = read.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # do we retain low-abundance components unchanged? if med < NORMALIZE_LIMIT and args.variable_coverage: write_record(read, corrfp) written_reads += 1 written_bp += len(read.sequence) skipped_n += 1 skipped_bp += len(read.sequence) # otherwise, examine/correct. else: # med >= NORMALIZE LIMIT or not args.variable_coverage is_aligned, new_seq = correct_sequence(aligner, seq) if is_aligned: if new_seq != read.sequence: corrected_reads += 1 read.sequence = new_seq if hasattr(read, 'quality'): fix_quality(read) write_record(read, corrfp) written_reads += 1 written_bp += len(new_seq) print('removing %s' % pass2filename, file=sys.stderr) os.unlink(pass2filename) print('removing temp directory & contents (%s)' % tempdir, file=sys.stderr) shutil.rmtree(tempdir) n_passes = 1.0 + (float(save_pass2_total) / n_reads) percent_reads_corrected = float(corrected_reads + (n_reads - written_reads)) /\ n_reads * 100.0 print('read %d reads, %d bp' % ( n_reads, n_bp, ), file=sys.stderr) print('wrote %d reads, %d bp' % ( written_reads, written_bp, ), file=sys.stderr) print('looked at %d reads twice (%.2f passes)' % (save_pass2_total, n_passes), file=sys.stderr) print('removed %d reads and corrected %d reads (%.2f%%)' % (n_reads - written_reads, corrected_reads, percent_reads_corrected), file=sys.stderr) print('removed %.2f%% of bases (%d total)' % ((1 - (written_bp / float(n_bp))) * 100.0, n_bp - written_bp), file=sys.stderr) if args.variable_coverage: percent_reads_hicov = 100.0 * float(n_reads - skipped_n) / n_reads print('%d reads were high coverage (%.2f%%);' % (n_reads - skipped_n, percent_reads_hicov), file=sys.stderr) print(('skipped %d reads/%d bases because of low coverage') % (skipped_n, skipped_bp), file=sys.stderr) fp_rate = \ khmer.calc_expected_collisions(ct, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print('fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate), file=sys.stderr) print('output in *.corr', file=sys.stderr) if args.savegraph: print("Saving k-mer countgraph to", args.savegraph, file=sys.stderr) ct.save(args.savegraph)
def main(): # pylint: disable=too-many-branches,too-many-statements info('normalize-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report force_single = args.force_single # check for similar filenames # if we're using a single output file only check for identical filenames # otherwise, check for identical BASE names as well. filenames = [] basenames = [] for pathfilename in args.input_filenames: filenames.append(pathfilename) if args.single_output_file: continue # nothing more to worry about basename = os.path.basename(pathfilename) if basename in basenames: print('ERROR: Duplicate filename--Cannot handle this!', file=sys.stderr) print('** Exiting!', file=sys.stderr) sys.exit(1) basenames.append(basename) # check that files exist and there is sufficient output disk space. check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable(args.n_tables * args.min_tablesize, args.force) # load or create counting table. if args.loadtable: print('loading k-mer counting table from ' + args.loadtable, file=sys.stderr) htable = khmer.load_counting_hash(args.loadtable) else: print('making k-mer counting table', file=sys.stderr) htable = khmer.new_counting_hash(args.ksize, args.min_tablesize, args.n_tables) input_filename = None # create an object to handle diginorm of all files norm = Normalizer(args.cutoff, htable) # make a list of all filenames and if they're paired or not; # if we don't know if they're paired, default to allowing but not # forcing pairing. files = [] for e in filenames: files.append([e, args.paired]) if args.unpaired_reads: files.append([args.unpaired_reads, False]) corrupt_files = [] outfp = None output_name = None if args.single_output_file: if args.single_output_file is sys.stdout: output_name = '/dev/stdout' else: output_name = args.single_output_file.name outfp = args.single_output_file # # main loop: iterate over all files given, do diginorm. # for filename, require_paired in files: if not args.single_output_file: output_name = os.path.basename(filename) + '.keep' outfp = open(output_name, 'w') # failsafe context manager in case an input file breaks with CatchIOErrors(filename, outfp, args.single_output_file, args.force, corrupt_files): screed_iter = screed.open(filename, parse_description=False) reader = broken_paired_reader(screed_iter, min_length=args.ksize, force_single=force_single, require_paired=require_paired) # actually do diginorm for record in WithDiagnostics(filename, norm, reader, report_fp): if record is not None: write_record(record, outfp) print('output in ' + output_name, file=sys.stderr) if output_name is not '/dev/stdout': outfp.close() # finished - print out some diagnostics. print('Total number of unique k-mers: {0}'.format(htable.n_unique_kmers()), file=sys.stderr) if args.savetable: print('...saving to ' + args.savetable, file=sys.stderr) htable.save(args.savetable) fp_rate = \ khmer.calc_expected_collisions(htable, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print('fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate), file=sys.stderr) if args.force and len(corrupt_files) > 0: print("** WARNING: Finished with errors!", file=sys.stderr) print("** IOErrors occurred in the following files:", file=sys.stderr) print("\t", " ".join(corrupt_files), file=sys.stderr)
def main(): # pylint: disable=too-many-branches,too-many-statements info('normalize-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable( args.n_tables * args.min_tablesize, args.force) # list to save error files along with throwing exceptions if args.force: corrupt_files = [] if args.loadtable: print 'loading k-mer counting table from', args.loadtable htable = khmer.load_counting_hash(args.loadtable) else: print 'making k-mer counting table' htable = khmer.new_counting_hash(args.ksize, args.min_tablesize, args.n_tables) total = 0 discarded = 0 input_filename = None if args.single_output_filename: output_name = args.single_output_filename if args.append: outfp = open(args.single_output_filename, 'a') else: outfp = open(args.single_output_filename, 'w') for index, input_filename in enumerate(args.input_filenames): if not args.single_output_filename: output_name = os.path.basename(input_filename) + '.keep' outfp = open(output_name, 'w') total_acc = 0 discarded_acc = 0 try: total_acc, discarded_acc = normalize_by_median(input_filename, outfp, htable, args, report_fp) except IOError as err: handle_error(err, output_name, input_filename, args.fail_save, htable) if not args.force: print >> sys.stderr, '** Exiting!' sys.exit(1) else: print >> sys.stderr, '*** Skipping error file, moving on...' corrupt_files.append(input_filename) else: if total_acc == 0 and discarded_acc == 0: print 'SKIPPED empty file', input_filename else: total += total_acc discarded += discarded_acc print 'DONE with {inp}; kept {kept} of {total} or {perc:2}%'\ .format(inp=input_filename, kept=total - discarded, total=total, perc=int(100. - discarded / float(total) * 100.)) print 'output in', output_name if (args.dump_frequency > 0 and index > 0 and index % args.dump_frequency == 0): print 'Backup: Saving k-mer counting file through', input_filename if args.savetable: hashname = args.savetable print '...saving to', hashname else: hashname = 'backup.ct' print 'Nothing given for savetable, saving to', hashname htable.save(hashname) if args.report_total_kmers: print >> sys.stderr, 'Total number of unique k-mers: {0}'.format( htable.n_unique_kmers()) if args.savetable: print 'Saving k-mer counting table through', input_filename print '...saving to', args.savetable htable.save(args.savetable) fp_rate = khmer.calc_expected_collisions(htable) print 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) if args.force and len(corrupt_files) > 0: print >> sys.stderr, "** WARNING: Finished with errors!" print >> sys.stderr, "** IOErrors occurred in the following files:" print >> sys.stderr, "\t", " ".join(corrupt_files) if fp_rate > MAX_FALSE_POSITIVE_RATE: print >> sys.stderr, "**" print >> sys.stderr, ("** ERROR: the k-mer counting table is too small" " for this data set. Increase tablesize/# " "tables.") print >> sys.stderr, "**" print >> sys.stderr, "** Do not use these results!!" if not args.force: sys.exit(1)
def main(): # pylint: disable=too-many-branches,too-many-statements info('normalize-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report # check for similar filenames filenames = [] for pathfilename in args.input_filenames: filename = pathfilename.split('/')[-1] if (filename in filenames): print >> sys.stderr, "WARNING: At least two input files are named \ %s . (The script normalize-by-median.py can not handle this, only one .keep \ file for one of the input files will be generated.)" % filename else: filenames.append(filename) # check for others check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable(args.n_tables * args.min_tablesize, args.force) # list to save error files along with throwing exceptions corrupt_files = [] if args.loadtable: print 'loading k-mer counting table from', args.loadtable htable = khmer.load_counting_hash(args.loadtable) else: print >> sys.stderr, 'making k-mer counting table' htable = khmer.new_counting_hash(args.ksize, args.min_tablesize, args.n_tables) input_filename = None for index, input_filename in enumerate(args.input_filenames): total_acc, discarded_acc, corrupt_files = \ normalize_by_median_and_check( input_filename, htable, args.single_output_file, args.fail_save, args.paired, args.cutoff, args.force, corrupt_files, report_fp) if (args.dump_frequency > 0 and index > 0 and index % args.dump_frequency == 0): print 'Backup: Saving k-mer counting file through', input_filename if args.savetable: hashname = args.savetable print '...saving to', hashname else: hashname = 'backup.ct' print 'Nothing given for savetable, saving to', hashname htable.save(hashname) if args.paired and args.unpaired_reads: args.paired = False output_name = args.unpaired_reads if not args.single_output_file: output_name = os.path.basename(args.unpaired_reads) + '.keep' outfp = open(output_name, 'w') total_acc, discarded_acc, corrupt_files = \ normalize_by_median_and_check( args.unpaired_reads, htable, args.single_output_file, args.fail_save, args.paired, args.cutoff, args.force, corrupt_files, report_fp) if args.report_total_kmers: print >> sys.stderr, 'Total number of unique k-mers: {0}'.format( htable.n_unique_kmers()) if args.savetable: print 'Saving k-mer counting table through', input_filename print '...saving to', args.savetable htable.save(args.savetable) fp_rate = \ khmer.calc_expected_collisions(htable, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print >> sys.stderr, \ 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) if args.force and len(corrupt_files) > 0: print >> sys.stderr, "** WARNING: Finished with errors!" print >> sys.stderr, "** IOErrors occurred in the following files:" print >> sys.stderr, "\t", " ".join(corrupt_files)
def main(): # pylint: disable=too-many-branches,too-many-statements info('saturate-by-median.py', ['diginorm']) parser = sanitize_help(get_parser()) args = parser.parse_args() report_on_config(args) report_fp = args.report report_frequency = args.report_frequency check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, False) if args.savegraph: check_space_for_graph(args, 'countgraph', False) # list to save error files along with throwing exceptions if args.force: corrupt_files = [] if args.loadgraph: print('loading k-mer countgraph from', args.loadgraph) htable = khmer.load_countgraph(args.loadgraph) else: print('making countgraph') htable = create_countgraph(args) total = 0 discarded = 0 for index, input_filename in enumerate(args.input_filenames): total_acc = 0 discarded_acc = 0 try: total_acc, discarded_acc = normalize_by_median(input_filename, htable, args, report_fp, report_frequency) except IOError as err: handle_error(err, input_filename) if not args.force: print("NOTE: This can be overridden using the --force" " argument", file=sys.stderr) print('** Exiting!', file=sys.stderr) sys.exit(1) else: print('*** Skipping error file, moving on...', file=sys.stderr) corrupt_files.append(input_filename) else: if total_acc == 0 and discarded_acc == 0: print('SKIPPED empty file', input_filename) else: total += total_acc discarded += discarded_acc print('DONE with {inp}; kept {kept} of {total} or {perc:2}%'\ .format(inp=input_filename, kept=total - discarded, total=total, perc=int(100. - discarded / float(total) * 100.))) if args.savegraph: print('Saving k-mer countgraph through', input_filename) print('...saving to', args.savegraph) htable.save(args.savegraph) # re: threshold, see Zhang et al., # http://arxiv.org/abs/1309.2975 fp_rate = khmer.calc_expected_collisions(htable, args.force, max_false_pos=.8) print('fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate)) if args.force and len(corrupt_files) > 0: print("** WARNING: Finished with errors!", file=sys.stderr) print("** I/O Errors occurred in the following files:", file=sys.stderr) print("\t", " ".join(corrupt_files), file=sys.stderr)
def main(): # pylint: disable=too-many-branches,too-many-statements info('saturate-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report report_frequency = args.report_frequency check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, False) if args.savetable: check_space_for_hashtable(args.n_tables * args.min_tablesize, False) # list to save error files along with throwing exceptions if args.force: corrupt_files = [] if args.loadtable: print 'loading k-mer counting table from', args.loadtable htable = khmer.load_counting_hash(args.loadtable) else: print 'making k-mer counting table' htable = khmer.new_counting_hash(args.ksize, args.min_tablesize, args.n_tables) total = 0 discarded = 0 for index, input_filename in enumerate(args.input_filenames): total_acc = 0 discarded_acc = 0 try: total_acc, discarded_acc = normalize_by_median(input_filename, htable, args, report_fp, report_frequency) except IOError as err: handle_error(err, input_filename) if not args.force: print >> sys.stderr, '** Exiting!' sys.exit(1) else: print >> sys.stderr, '*** Skipping error file, moving on...' corrupt_files.append(input_filename) else: if total_acc == 0 and discarded_acc == 0: print 'SKIPPED empty file', input_filename else: total += total_acc discarded += discarded_acc print 'DONE with {inp}; kept {kept} of {total} or {perc:2}%'\ .format(inp=input_filename, kept=total - discarded, total=total, perc=int(100. - discarded / float(total) * 100.)) if args.savetable: print 'Saving k-mer counting table through', input_filename print '...saving to', args.savetable htable.save(args.savetable) # re: threshold, see Zhang et al., # http://arxiv.org/abs/1309.2975 fp_rate = khmer.calc_expected_collisions(htable, args.force, max_false_pos=.8) print 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) if args.force and len(corrupt_files) > 0: print >> sys.stderr, "** WARNING: Finished with errors!" print >> sys.stderr, "** IOErrors occurred in the following files:" print >> sys.stderr, "\t", " ".join(corrupt_files)
def main(): # pylint: disable=too-many-branches,too-many-statements info('saturate-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report report_frequency = args.report_frequency check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, False) if args.savetable: check_space_for_hashtable(args.n_tables * args.min_tablesize, False) # list to save error files along with throwing exceptions if args.force: corrupt_files = [] if args.loadtable: print 'loading k-mer counting table from', args.loadtable htable = khmer.load_counting_hash(args.loadtable) else: print 'making k-mer counting table' htable = khmer.new_counting_hash(args.ksize, args.min_tablesize, args.n_tables) total = 0 discarded = 0 for index, input_filename in enumerate(args.input_filenames): total_acc = 0 discarded_acc = 0 try: total_acc, discarded_acc = normalize_by_median( input_filename, htable, args, report_fp, report_frequency) except IOError as err: handle_error(err, input_filename) if not args.force: print >> sys.stderr, '** Exiting!' sys.exit(1) else: print >> sys.stderr, '*** Skipping error file, moving on...' corrupt_files.append(input_filename) else: if total_acc == 0 and discarded_acc == 0: print 'SKIPPED empty file', input_filename else: total += total_acc discarded += discarded_acc print 'DONE with {inp}; kept {kept} of {total} or {perc:2}%'\ .format(inp=input_filename, kept=total - discarded, total=total, perc=int(100. - discarded / float(total) * 100.)) if args.savetable: print 'Saving k-mer counting table through', input_filename print '...saving to', args.savetable htable.save(args.savetable) # re: threshold, see Zhang et al., # http://arxiv.org/abs/1309.2975 fp_rate = khmer.calc_expected_collisions(htable, args.force, max_false_pos=.8) print 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) if args.force and len(corrupt_files) > 0: print >> sys.stderr, "** WARNING: Finished with errors!" print >> sys.stderr, "** IOErrors occurred in the following files:" print >> sys.stderr, "\t", " ".join(corrupt_files)
def main(): info('trim-low-abund.py', ['streaming']) parser = get_parser() args = parser.parse_args() ### if len(set(args.input_filenames)) != len(args.input_filenames): print >>sys.stderr, \ "Error: Cannot input the same filename multiple times." sys.exit(1) ### report_on_config(args) check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable( args.n_tables * args.min_tablesize, args.force) K = args.ksize CUTOFF = args.cutoff NORMALIZE_LIMIT = args.normalize_to if args.loadtable: print >>sys.stderr, 'loading k-mer counting table from', args.loadtable ct = khmer.load_counting_hash(args.loadtable) else: print >>sys.stderr, 'making k-mer counting table' ct = khmer.new_counting_hash(K, args.min_tablesize, args.n_tables) tempdir = tempfile.mkdtemp('khmer', 'tmp', args.tempdir) print >>sys.stderr, 'created temporary directory %s; ' \ 'use -T to change location' % tempdir # ### FIRST PASS ### save_pass2_total = 0 n_bp = 0 n_reads = 0 written_bp = 0 written_reads = 0 trimmed_reads = 0 pass2list = [] for filename in args.input_filenames: pass2filename = os.path.basename(filename) + '.pass2' pass2filename = os.path.join(tempdir, pass2filename) trimfilename = os.path.basename(filename) + '.abundtrim' pass2list.append((filename, pass2filename, trimfilename)) screed_iter = screed.open(filename, parse_description=False) pass2fp = open(pass2filename, 'w') trimfp = open(trimfilename, 'w') save_pass2 = 0 n = 0 paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=args.ignore_pairs) for n, is_pair, read1, read2 in paired_iter: if n % 10000 == 0: print >>sys.stderr, '...', n, filename, save_pass2, \ n_reads, n_bp, written_reads, written_bp # we want to track paired reads here, to make sure that pairs # are not split between first pass and second pass. if is_pair: n_reads += 2 n_bp += len(read1.sequence) + len(read2.sequence) seq1 = read1.sequence.replace('N', 'A') seq2 = read2.sequence.replace('N', 'A') med1, _, _ = ct.get_median_count(seq1) med2, _, _ = ct.get_median_count(seq2) if med1 < NORMALIZE_LIMIT or med2 < NORMALIZE_LIMIT: ct.consume(seq1) ct.consume(seq2) write_record_pair(read1, read2, pass2fp) save_pass2 += 2 else: _, trim_at1 = ct.trim_on_abundance(seq1, CUTOFF) _, trim_at2 = ct.trim_on_abundance(seq2, CUTOFF) if trim_at1 >= K: read1 = trim_record(read1, trim_at1) if trim_at2 >= K: read2 = trim_record(read2, trim_at2) if trim_at1 != len(seq1): trimmed_reads += 1 if trim_at2 != len(seq2): trimmed_reads += 1 write_record_pair(read1, read2, trimfp) written_reads += 2 written_bp += trim_at1 + trim_at2 else: n_reads += 1 n_bp += len(read1.sequence) seq = read1.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # has this portion of the graph saturated? if not, # consume & save => pass2. if med < NORMALIZE_LIMIT: ct.consume(seq) write_record(read1, pass2fp) save_pass2 += 1 else: # trim!! _, trim_at = ct.trim_on_abundance(seq, CUTOFF) if trim_at >= K: new_read = trim_record(read1, trim_at) write_record(new_read, trimfp) written_reads += 1 written_bp += trim_at if trim_at != len(read1.sequence): trimmed_reads += 1 pass2fp.close() trimfp.close() print '%s: kept aside %d of %d from first pass, in %s' % \ (filename, save_pass2, n, filename) save_pass2_total += save_pass2 # ### SECOND PASS. ### skipped_n = 0 skipped_bp = 0 for _, pass2filename, trimfilename in pass2list: print 'second pass: looking at sequences kept aside in %s' % \ pass2filename # note that for this second pass, we don't care about paired # reads - they will be output in the same order they're read in, # so pairs will stay together if not orphaned. This is in contrast # to the first loop. trimfp = open(trimfilename, 'a') for n, read in enumerate(screed.open(pass2filename, parse_description=False)): if n % 10000 == 0: print >>sys.stderr, '... x 2', n, pass2filename, \ written_reads, written_bp seq = read.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # do we retain low-abundance components unchanged? if med < NORMALIZE_LIMIT and args.variable_coverage: write_record(read, trimfp) written_reads += 1 written_bp += len(read.sequence) skipped_n += 1 skipped_bp += len(read.sequence) # otherwise, examine/trim/truncate. else: # med >= NORMALIZE LIMIT or not args.variable_coverage _, trim_at = ct.trim_on_abundance(seq, CUTOFF) if trim_at >= K: new_read = trim_record(read, trim_at) write_record(new_read, trimfp) written_reads += 1 written_bp += trim_at if trim_at != len(read.sequence): trimmed_reads += 1 print >>sys.stderr, 'removing %s' % pass2filename os.unlink(pass2filename) print >>sys.stderr, 'removing temp directory & contents (%s)' % tempdir shutil.rmtree(tempdir) n_passes = 1.0 + (float(save_pass2_total) / n_reads) percent_reads_trimmed = float(trimmed_reads + (n_reads - written_reads)) /\ n_reads * 100.0 print 'read %d reads, %d bp' % (n_reads, n_bp,) print 'wrote %d reads, %d bp' % (written_reads, written_bp,) print 'looked at %d reads twice (%.2f passes)' % (save_pass2_total, n_passes) print 'removed %d reads and trimmed %d reads (%.2f%%)' % \ (n_reads - written_reads, trimmed_reads, percent_reads_trimmed) print 'trimmed or removed %.2f%% of bases (%d total)' % \ ((1 - (written_bp / float(n_bp))) * 100.0, n_bp - written_bp) if args.variable_coverage: percent_reads_hicov = 100.0 * float(n_reads - skipped_n) / n_reads print '%d reads were high coverage (%.2f%%);' % (n_reads - skipped_n, percent_reads_hicov) print 'skipped %d reads/%d bases because of low coverage' % \ (skipped_n, skipped_bp) fp_rate = khmer.calc_expected_collisions(ct) print >>sys.stderr, \ 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) if fp_rate > MAX_FALSE_POSITIVE_RATE: print >> sys.stderr, "**" print >> sys.stderr, ("** ERROR: the k-mer counting table is too small" " for this data set. Increase tablesize/# " "tables.") print >> sys.stderr, "**" print >> sys.stderr, "** Do not use these results!!" sys.exit(1) print 'output in *.abundtrim' if args.savetable: print >>sys.stderr, "Saving k-mer counting table to", args.savetable ct.save(args.savetable)
def main(): info('trim-low-abund.py', ['streaming']) parser = get_parser() args = parser.parse_args() ### if len(set(args.input_filenames)) != len(args.input_filenames): print >>sys.stderr, \ "Error: Cannot input the same filename multiple times." sys.exit(1) ### report_on_config(args) check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable(args.n_tables * args.min_tablesize, args.force) K = args.ksize CUTOFF = args.cutoff NORMALIZE_LIMIT = args.normalize_to if args.loadtable: print >> sys.stderr, 'loading k-mer counting table from', args.loadtable ct = khmer.load_counting_hash(args.loadtable) else: print >> sys.stderr, 'making k-mer counting table' ct = khmer.new_counting_hash(K, args.min_tablesize, args.n_tables) tempdir = tempfile.mkdtemp('khmer', 'tmp', args.tempdir) print >>sys.stderr, 'created temporary directory %s; ' \ 'use -T to change location' % tempdir # ### FIRST PASS ### save_pass2_total = 0 n_bp = 0 n_reads = 0 written_bp = 0 written_reads = 0 trimmed_reads = 0 pass2list = [] for filename in args.input_filenames: pass2filename = os.path.basename(filename) + '.pass2' pass2filename = os.path.join(tempdir, pass2filename) if args.out is None: trimfp = open(os.path.basename(filename) + '.abundtrim', 'w') else: trimfp = args.out pass2list.append((filename, pass2filename, trimfp)) screed_iter = screed.open(filename, parse_description=False) pass2fp = open(pass2filename, 'w') save_pass2 = 0 n = 0 paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=args.ignore_pairs) for n, is_pair, read1, read2 in paired_iter: if n % 10000 == 0: print >>sys.stderr, '...', n, filename, save_pass2, \ n_reads, n_bp, written_reads, written_bp # we want to track paired reads here, to make sure that pairs # are not split between first pass and second pass. if is_pair: n_reads += 2 n_bp += len(read1.sequence) + len(read2.sequence) seq1 = read1.sequence.replace('N', 'A') seq2 = read2.sequence.replace('N', 'A') med1, _, _ = ct.get_median_count(seq1) med2, _, _ = ct.get_median_count(seq2) if med1 < NORMALIZE_LIMIT or med2 < NORMALIZE_LIMIT: ct.consume(seq1) ct.consume(seq2) write_record_pair(read1, read2, pass2fp) save_pass2 += 2 else: _, trim_at1 = ct.trim_on_abundance(seq1, CUTOFF) _, trim_at2 = ct.trim_on_abundance(seq2, CUTOFF) if trim_at1 >= K: read1 = trim_record(read1, trim_at1) if trim_at2 >= K: read2 = trim_record(read2, trim_at2) if trim_at1 != len(seq1): trimmed_reads += 1 if trim_at2 != len(seq2): trimmed_reads += 1 write_record_pair(read1, read2, trimfp) written_reads += 2 written_bp += trim_at1 + trim_at2 else: n_reads += 1 n_bp += len(read1.sequence) seq = read1.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # has this portion of the graph saturated? if not, # consume & save => pass2. if med < NORMALIZE_LIMIT: ct.consume(seq) write_record(read1, pass2fp) save_pass2 += 1 else: # trim!! _, trim_at = ct.trim_on_abundance(seq, CUTOFF) if trim_at >= K: new_read = trim_record(read1, trim_at) write_record(new_read, trimfp) written_reads += 1 written_bp += trim_at if trim_at != len(read1.sequence): trimmed_reads += 1 pass2fp.close() print >>sys.stderr, '%s: kept aside %d of %d from first pass, in %s' \ % (filename, save_pass2, n, filename) save_pass2_total += save_pass2 # ### SECOND PASS. ### skipped_n = 0 skipped_bp = 0 for _, pass2filename, trimfp in pass2list: print >> sys.stderr, ('second pass: looking at sequences kept aside ' 'in %s') % pass2filename # note that for this second pass, we don't care about paired # reads - they will be output in the same order they're read in, # so pairs will stay together if not orphaned. This is in contrast # to the first loop. for n, read in enumerate( screed.open(pass2filename, parse_description=False)): if n % 10000 == 0: print >>sys.stderr, '... x 2', n, pass2filename, \ written_reads, written_bp seq = read.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # do we retain low-abundance components unchanged? if med < NORMALIZE_LIMIT and args.variable_coverage: write_record(read, trimfp) written_reads += 1 written_bp += len(read.sequence) skipped_n += 1 skipped_bp += len(read.sequence) # otherwise, examine/trim/truncate. else: # med >= NORMALIZE LIMIT or not args.variable_coverage _, trim_at = ct.trim_on_abundance(seq, CUTOFF) if trim_at >= K: new_read = trim_record(read, trim_at) write_record(new_read, trimfp) written_reads += 1 written_bp += trim_at if trim_at != len(read.sequence): trimmed_reads += 1 print >> sys.stderr, 'removing %s' % pass2filename os.unlink(pass2filename) print >> sys.stderr, 'removing temp directory & contents (%s)' % tempdir shutil.rmtree(tempdir) n_passes = 1.0 + (float(save_pass2_total) / n_reads) percent_reads_trimmed = float(trimmed_reads + (n_reads - written_reads)) /\ n_reads * 100.0 print >> sys.stderr, 'read %d reads, %d bp' % ( n_reads, n_bp, ) print >> sys.stderr, 'wrote %d reads, %d bp' % ( written_reads, written_bp, ) print >>sys.stderr, 'looked at %d reads twice (%.2f passes)' % \ (save_pass2_total, n_passes) print >>sys.stderr, 'removed %d reads and trimmed %d reads (%.2f%%)' % \ (n_reads - written_reads, trimmed_reads, percent_reads_trimmed) print >>sys.stderr, 'trimmed or removed %.2f%% of bases (%d total)' % \ ((1 - (written_bp / float(n_bp))) * 100.0, n_bp - written_bp) if args.variable_coverage: percent_reads_hicov = 100.0 * float(n_reads - skipped_n) / n_reads print >>sys.stderr, '%d reads were high coverage (%.2f%%);' % \ (n_reads - skipped_n, percent_reads_hicov) print >> sys.stderr, ('skipped %d reads/%d bases because of low' 'coverage') % (skipped_n, skipped_bp) fp_rate = \ khmer.calc_expected_collisions(ct, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print >>sys.stderr, \ 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) print >> sys.stderr, 'output in *.abundtrim' if args.savetable: print >> sys.stderr, "Saving k-mer counting table to", args.savetable ct.save(args.savetable)
def main(): info('trim-low-abund.py', ['streaming']) parser = sanitize_help(get_parser()) args = parser.parse_args() ### if len(set(args.input_filenames)) != len(args.input_filenames): print("Error: Cannot input the same filename multiple times.", file=sys.stderr) sys.exit(1) ### report_on_config(args) check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savegraph: graphsize = calculate_graphsize(args, 'countgraph') check_space_for_graph(args.savegraph, graphsize, args.force) if ('-' in args.input_filenames or '/dev/stdin' in args.input_filenames) \ and not args.output: print("Accepting input from stdin; output filename must " "be provided with -o.", file=sys.stderr) sys.exit(1) if args.loadgraph: print('loading countgraph from', args.loadgraph, file=sys.stderr) ct = khmer.load_countgraph(args.loadgraph) else: print('making countgraph', file=sys.stderr) ct = khmer_args.create_countgraph(args) K = ct.ksize() CUTOFF = args.cutoff NORMALIZE_LIMIT = args.normalize_to tempdir = tempfile.mkdtemp('khmer', 'tmp', args.tempdir) print('created temporary directory %s; ' 'use -T to change location' % tempdir, file=sys.stderr) # ### FIRST PASS ### save_pass2_total = 0 n_bp = 0 n_reads = 0 written_bp = 0 written_reads = 0 trimmed_reads = 0 pass2list = [] for filename in args.input_filenames: pass2filename = os.path.basename(filename) + '.pass2' pass2filename = os.path.join(tempdir, pass2filename) if args.output is None: trimfp = get_file_writer(open(os.path.basename(filename) + '.abundtrim', 'wb'), args.gzip, args.bzip) else: trimfp = get_file_writer(args.output, args.gzip, args.bzip) pass2list.append((filename, pass2filename, trimfp)) screed_iter = screed.open(filename) pass2fp = open(pass2filename, 'w') save_pass2 = 0 n = 0 paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=args.ignore_pairs) for n, is_pair, read1, read2 in paired_iter: if n % 10000 == 0: print('...', n, filename, save_pass2, n_reads, n_bp, written_reads, written_bp, file=sys.stderr) # we want to track paired reads here, to make sure that pairs # are not split between first pass and second pass. if is_pair: n_reads += 2 n_bp += len(read1.sequence) + len(read2.sequence) seq1 = read1.sequence.replace('N', 'A') seq2 = read2.sequence.replace('N', 'A') med1, _, _ = ct.get_median_count(seq1) med2, _, _ = ct.get_median_count(seq2) if med1 < NORMALIZE_LIMIT or med2 < NORMALIZE_LIMIT: ct.consume(seq1) ct.consume(seq2) write_record_pair(read1, read2, pass2fp) save_pass2 += 2 else: _, trim_at1 = ct.trim_on_abundance(seq1, CUTOFF) _, trim_at2 = ct.trim_on_abundance(seq2, CUTOFF) if trim_at1 >= K: read1 = trim_record(read1, trim_at1) if trim_at2 >= K: read2 = trim_record(read2, trim_at2) if trim_at1 != len(seq1): trimmed_reads += 1 if trim_at2 != len(seq2): trimmed_reads += 1 write_record_pair(read1, read2, trimfp) written_reads += 2 written_bp += trim_at1 + trim_at2 else: n_reads += 1 n_bp += len(read1.sequence) seq = read1.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # has this portion of the graph saturated? if not, # consume & save => pass2. if med < NORMALIZE_LIMIT: ct.consume(seq) write_record(read1, pass2fp) save_pass2 += 1 else: # trim!! _, trim_at = ct.trim_on_abundance(seq, CUTOFF) if trim_at >= K: new_read = trim_record(read1, trim_at) write_record(new_read, trimfp) written_reads += 1 written_bp += trim_at if trim_at != len(read1.sequence): trimmed_reads += 1 pass2fp.close() print('%s: kept aside %d of %d from first pass, in %s' % (filename, save_pass2, n, filename), file=sys.stderr) save_pass2_total += save_pass2 # ### SECOND PASS. ### skipped_n = 0 skipped_bp = 0 for _, pass2filename, trimfp in pass2list: print('second pass: looking at sequences kept aside in %s' % pass2filename, file=sys.stderr) # note that for this second pass, we don't care about paired # reads - they will be output in the same order they're read in, # so pairs will stay together if not orphaned. This is in contrast # to the first loop. for n, read in enumerate(screed.open(pass2filename)): if n % 10000 == 0: print('... x 2', n, pass2filename, written_reads, written_bp, file=sys.stderr) seq = read.sequence.replace('N', 'A') med, _, _ = ct.get_median_count(seq) # do we retain low-abundance components unchanged? if med < NORMALIZE_LIMIT and args.variable_coverage: write_record(read, trimfp) written_reads += 1 written_bp += len(read.sequence) skipped_n += 1 skipped_bp += len(read.sequence) # otherwise, examine/trim/truncate. else: # med >= NORMALIZE LIMIT or not args.variable_coverage _, trim_at = ct.trim_on_abundance(seq, CUTOFF) if trim_at >= K: new_read = trim_record(read, trim_at) write_record(new_read, trimfp) written_reads += 1 written_bp += trim_at if trim_at != len(read.sequence): trimmed_reads += 1 print('removing %s' % pass2filename, file=sys.stderr) os.unlink(pass2filename) print('removing temp directory & contents (%s)' % tempdir, file=sys.stderr) shutil.rmtree(tempdir) n_passes = 1.0 + (float(save_pass2_total) / n_reads) percent_reads_trimmed = float(trimmed_reads + (n_reads - written_reads)) /\ n_reads * 100.0 print('read %d reads, %d bp' % (n_reads, n_bp,), file=sys.stderr) print('wrote %d reads, %d bp' % (written_reads, written_bp,), file=sys.stderr) print('looked at %d reads twice (%.2f passes)' % (save_pass2_total, n_passes), file=sys.stderr) print('removed %d reads and trimmed %d reads (%.2f%%)' % (n_reads - written_reads, trimmed_reads, percent_reads_trimmed), file=sys.stderr) print('trimmed or removed %.2f%% of bases (%d total)' % ((1 - (written_bp / float(n_bp))) * 100.0, n_bp - written_bp), file=sys.stderr) if args.variable_coverage: percent_reads_hicov = 100.0 * float(n_reads - skipped_n) / n_reads print('%d reads were high coverage (%.2f%%);' % (n_reads - skipped_n, percent_reads_hicov), file=sys.stderr) print('skipped %d reads/%d bases because of low coverage' % (skipped_n, skipped_bp), file=sys.stderr) fp_rate = \ khmer.calc_expected_collisions(ct, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print('fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate), file=sys.stderr) print('output in *.abundtrim', file=sys.stderr) if args.savegraph: print("Saving k-mer countgraph to", args.savegraph, file=sys.stderr) ct.save(args.savegraph)
def main(): info("sweep-reads-buffered.py", ["sweep"]) parser = sanitize_epilog(get_parser()) args = parser.parse_args() if args.max_tablesize < MAX_HSIZE: args.max_tablesize = MAX_HSIZE if args.ksize < MIN_KSIZE: args.ksize = MIN_KSIZE report_on_config(args, graphtype="nodegraph") K = args.ksize HT_SIZE = args.max_tablesize N_HT = args.n_tables traversal_range = args.traversal_range input_fastp = args.input_fastp if not args.outdir: outdir = os.path.dirname(input_fastp) else: outdir = args.outdir max_buffers = args.max_buffers output_pref = args.output_prefix buf_size = args.buffer_size max_reads = args.max_reads check_input_files(args.input_fastp, args.force) check_valid_file_exists(args.input_files) all_input_files = [input_fastp] all_input_files.extend(args.input_files) # Check disk space availability check_space(all_input_files, args.force) # figure out input file type (FA/FQ) -- based on first file ix = iter(screed.open(args.input_files[0])) record = next(ix) del ix extension = "fa" if hasattr(record, "quality"): # fastq! extension = "fq" output_buffer = ReadBufferManager(max_buffers, max_reads, buf_size, output_pref, outdir, extension) # consume the partitioned fasta with which to label the graph ht = khmer.GraphLabels(K, HT_SIZE, N_HT) try: print("consuming input sequences...", file=sys.stderr) if args.label_by_pid: print("...labeling by partition id (pid)", file=sys.stderr) ht.consume_partitioned_fasta_and_tag_with_labels(input_fastp) elif args.label_by_seq: print("...labeling by sequence", file=sys.stderr) for n, record in enumerate(screed.open(input_fastp)): if n % 50000 == 0: print("...consumed {n} sequences...".format(n=n), file=sys.stderr) ht.consume_sequence_and_tag_with_labels(record.sequence, n) else: print("...labeling to create groups of size {s}".format(s=args.group_size), file=sys.stderr) label = -1 g = 0 try: outfp = open("{pref}_base_{g}.{ext}".format(pref=output_pref, g=g, ext=extension), "wb") for n, record in enumerate(screed.open(input_fastp)): if n % args.group_size == 0: label += 1 if label > g: g = label outfp = open("{pref}_base_{g}.{ext}".format(pref=output_pref, g=g, ext=extension), "wb") if n % 50000 == 0: print("...consumed {n} sequences...".format(n=n), file=sys.stderr) ht.consume_sequence_and_tag_with_labels(record.sequence, label) write_record(record, outfp) except (IOError, OSError) as e: print("!! ERROR !!", e, file=sys.stderr) print("...error splitting input. exiting...", file=sys.stderr) except (IOError, OSError) as e: print("!! ERROR: !!", e, file=sys.stderr) print( "...error consuming \ {i}. exiting...".format( i=input_fastp ), file=sys.stderr, ) print( "done consuming input sequence. \ added {t} tags and {l} \ labels...".format( t=ht.graph.n_tags(), l=ht.n_labels() ) ) label_dict = defaultdict(int) label_number_dist = [] n_orphaned = 0 n_labeled = 0 n_mlabeled = 0 total_t = time.clock() start_t = time.clock() for read_file in args.input_files: print("** sweeping {read_file} for labels...".format(read_file=read_file), file=sys.stderr) file_t = 0.0 try: read_fp = screed.open(read_file) except (IOError, OSError) as error: print("!! ERROR: !!", error, file=sys.stderr) print("*** Could not open {fn}, skipping...".format(fn=read_file), file=sys.stderr) else: for _, record in enumerate(read_fp): if _ % 50000 == 0: end_t = time.clock() batch_t = end_t - start_t file_t += batch_t print( "\tswept {n} reads [{nc} labeled, \ {no} orphaned] \ ** {sec}s ({sect}s total)".format( n=_, nc=n_labeled, no=n_orphaned, sec=batch_t, sect=file_t ), file=sys.stderr, ) start_t = time.clock() seq = record.sequence name = record.name try: labels = ht.sweep_label_neighborhood(seq, traversal_range) except ValueError as e: pass else: if hasattr(record, "quality"): seq_str = fmt_fastq(name, seq, record.quality, labels) else: seq_str = fmt_fasta(name, seq, labels) label_number_dist.append(len(labels)) if labels: n_labeled += 1 if len(labels) > 1: output_buffer.queue(seq_str, "multi") n_mlabeled += 1 label_dict["multi"] += 1 else: output_buffer.queue(seq_str, labels[0]) label_dict[labels[0]] += 1 else: n_orphaned += 1 output_buffer.queue(seq_str, "orphaned") label_dict["orphaned"] += 1 print("** End of file {fn}...".format(fn=read_file), file=sys.stderr) output_buffer.flush_all() read_fp.close() # gotta output anything left in the buffers at the end! print("** End of run...", file=sys.stderr) output_buffer.flush_all() total_t = time.clock() - total_t if output_buffer.num_write_errors > 0 or output_buffer.num_file_errors > 0: print("! WARNING: Sweep finished with errors !", file=sys.stderr) print("** {writee} reads not written".format(writee=output_buffer.num_write_errors), file=sys.stderr) print("** {filee} errors opening files".format(filee=output_buffer.num_file_errors), file=sys.stderr) print("swept {n_reads} for labels...".format(n_reads=n_labeled + n_orphaned), file=sys.stderr) print("...with {nc} labeled and {no} orphaned".format(nc=n_labeled, no=n_orphaned), file=sys.stderr) print("...and {nmc} multilabeled".format(nmc=n_mlabeled), file=sys.stderr) print("** outputting label number distribution...", file=sys.stderr) fn = os.path.join(outdir, "{pref}.dist.txt".format(pref=output_pref)) with open(fn, "w", encoding="utf-8") as outfp: for nc in label_number_dist: outfp.write("{nc}\n".format(nc=nc)) fn = os.path.join(outdir, "{pref}.counts.csv".format(pref=output_pref)) print("** outputting label read counts...", file=sys.stderr) with open(fn, "w", encoding="utf-8") as outfp: for k in label_dict: outfp.write("{l},{c}\n".format(l=k, c=label_dict[k]))
def main(): info('sweep-reads-buffered.py', ['sweep']) parser = get_parser() args = parser.parse_args() if args.min_tablesize < MIN_HSIZE: args.min_tablesize = MIN_HSIZE if args.ksize < MIN_KSIZE: args.ksize = MIN_KSIZE report_on_config(args, hashtype='hashbits') K = args.ksize HT_SIZE = args.min_tablesize N_HT = args.n_tables traversal_range = args.traversal_range input_fastp = args.input_fastp if not args.outdir: outdir = os.path.dirname(input_fastp) else: outdir = args.outdir max_buffers = args.max_buffers output_pref = args.output_prefix buf_size = args.buffer_size max_reads = args.max_reads check_file_status(args.input_fastp, args.force) check_valid_file_exists(args.input_files) all_input_files = [input_fastp] all_input_files.extend(args.input_files) # Check disk space availability check_space(all_input_files, args.force) # figure out input file type (FA/FQ) -- based on first file ix = iter(screed.open(args.input_files[0])) record = ix.next() del ix extension = 'fa' if hasattr(record, 'accuracy'): # fastq! extension = 'fq' output_buffer = ReadBufferManager( max_buffers, max_reads, buf_size, output_pref, outdir, extension) # consume the partitioned fasta with which to label the graph ht = khmer.LabelHash(K, HT_SIZE, N_HT) try: print >>sys.stderr, 'consuming input sequences...' if args.label_by_pid: print >>sys.stderr, '...labeling by partition id (pid)' ht.consume_partitioned_fasta_and_tag_with_labels(input_fastp) elif args.label_by_seq: print >>sys.stderr, '...labeling by sequence' for n, record in enumerate(screed.open(input_fastp)): if n % 50000 == 0: print >>sys.stderr, \ '...consumed {n} sequences...'.format(n=n) ht.consume_sequence_and_tag_with_labels(record.sequence, n) else: print >>sys.stderr, \ '...labeling to create groups of size {s}'.format( s=args.group_size) label = -1 g = 0 try: outfp = open('{pref}_base_{g}.{ext}'.format(pref=output_pref, g=g, ext=extension ), 'wb') for n, record in enumerate(screed.open(input_fastp)): if n % args.group_size == 0: label += 1 if label > g: g = label outfp = open('{pref}_base_{g}.{ext}'.format( pref=output_pref, g=g, ext=extension), 'wb') if n % 50000 == 0: print >>sys.stderr, \ '...consumed {n} sequences...'.format(n=n) ht.consume_sequence_and_tag_with_labels(record.sequence, label) if hasattr(record, 'accuracy'): outfp.write('@{name}\n{seq}+{accuracy}\n'.format( name=record.name, seq=record.sequence, accuracy=record.accuracy)) else: outfp.write('>{name}\n{seq}\n'.format( name=record.name, seq=record.sequence)) except IOError as e: print >>sys.stderr, '!! ERROR !!', e print >>sys.stderr, '...error splitting input. exiting...' except IOError as e: print >>sys.stderr, '!! ERROR: !!', e print >>sys.stderr, '...error consuming \ {i}. exiting...'.format(i=input_fastp) print >>sys.stderr, 'done consuming input sequence. \ added {t} tags and {l} \ labels...'.format(t=ht.n_tags(), l=ht.n_labels()) label_dict = defaultdict(int) label_number_dist = [] n_orphaned = 0 n_labeled = 0 n_mlabeled = 0 total_t = time.clock() start_t = time.clock() for read_file in args.input_files: print >>sys.stderr, '** sweeping {read_file} for labels...'.format( read_file=read_file) file_t = 0.0 try: read_fp = screed.open(read_file) except IOError as error: print >>sys.stderr, '!! ERROR: !!', error print >>sys.stderr, '*** Could not open {fn}, skipping...'.format( fn=read_file) else: for _, record in enumerate(read_fp): if _ % 50000 == 0: end_t = time.clock() batch_t = end_t - start_t file_t += batch_t print >>sys.stderr, '\tswept {n} reads [{nc} labeled, \ {no} orphaned] \ ** {sec}s ({sect}s total)' \ .format(n=_, nc=n_labeled, no=n_orphaned, sec=batch_t, sect=file_t) start_t = time.clock() seq = record.sequence name = record.name try: labels = ht.sweep_label_neighborhood(seq, traversal_range) except ValueError as e: pass else: if hasattr(record, 'accuracy'): seq_str = fmt_fastq(name, seq, record.accuracy, labels) else: seq_str = fmt_fasta(name, seq, labels) label_number_dist.append(len(labels)) if labels: n_labeled += 1 if len(labels) > 1: output_buffer.queue(seq_str, 'multi') n_mlabeled += 1 label_dict['multi'] += 1 else: output_buffer.queue(seq_str, labels[0]) label_dict[labels[0]] += 1 else: n_orphaned += 1 output_buffer.queue(seq_str, 'orphaned') label_dict['orphaned'] += 1 print >>sys.stderr, '** End of file {fn}...'.format(fn=read_file) output_buffer.flush_all() read_fp.close() # gotta output anything left in the buffers at the end! print >>sys.stderr, '** End of run...' output_buffer.flush_all() total_t = time.clock() - total_t if output_buffer.num_write_errors > 0 or output_buffer.num_file_errors > 0: print >>sys.stderr, '! WARNING: Sweep finished with errors !' print >>sys.stderr, '** {writee} reads not written'.format( writee=output_buffer.num_write_errors) print >>sys.stderr, '** {filee} errors opening files'.format( filee=output_buffer.num_file_errors) print >>sys.stderr, 'swept {n_reads} for labels...'.format( n_reads=n_labeled + n_orphaned) print >>sys.stderr, '...with {nc} labeled and {no} orphaned'.format( nc=n_labeled, no=n_orphaned) print >>sys.stderr, '...and {nmc} multilabeled'.format(nmc=n_mlabeled) print >>sys.stderr, '** outputting label number distribution...' fn = os.path.join(outdir, '{pref}.dist.txt'.format(pref=output_pref)) with open(fn, 'wb') as outfp: for nc in label_number_dist: outfp.write('{nc}\n'.format(nc=nc)) fn = os.path.join(outdir, '{pref}.counts.csv'.format(pref=output_pref)) print >>sys.stderr, '** outputting label read counts...' with open(fn, 'wb') as outfp: for k in label_dict: outfp.write('{l},{c}\n'.format(l=k, c=label_dict[k]))
def main(): # pylint: disable=too-many-branches,too-many-statements info('normalize-by-median.py', ['diginorm']) args = get_parser().parse_args() report_on_config(args) report_fp = args.report # check for similar filenames filenames = [] for pathfilename in args.input_filenames: filename = pathfilename.split('/')[-1] if (filename in filenames): print >>sys.stderr, "WARNING: At least two input files are named \ %s . (The script normalize-by-median.py can not handle this, only one .keep \ file for one of the input files will be generated.)" % filename else: filenames.append(filename) # check for others check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savetable: check_space_for_hashtable( args.n_tables * args.min_tablesize, args.force) # list to save error files along with throwing exceptions corrupt_files = [] if args.loadtable: print 'loading k-mer counting table from', args.loadtable htable = khmer.load_counting_hash(args.loadtable) else: print >> sys.stderr, 'making k-mer counting table' htable = khmer.new_counting_hash(args.ksize, args.min_tablesize, args.n_tables) input_filename = None for index, input_filename in enumerate(args.input_filenames): total_acc, discarded_acc, corrupt_files = \ normalize_by_median_and_check( input_filename, htable, args.single_output_file, args.fail_save, args.paired, args.cutoff, args.force, corrupt_files, report_fp) if (args.dump_frequency > 0 and index > 0 and index % args.dump_frequency == 0): print 'Backup: Saving k-mer counting file through', input_filename if args.savetable: hashname = args.savetable print '...saving to', hashname else: hashname = 'backup.ct' print 'Nothing given for savetable, saving to', hashname htable.save(hashname) if args.paired and args.unpaired_reads: args.paired = False output_name = args.unpaired_reads if not args.single_output_file: output_name = os.path.basename(args.unpaired_reads) + '.keep' outfp = open(output_name, 'w') total_acc, discarded_acc, corrupt_files = \ normalize_by_median_and_check( args.unpaired_reads, htable, args.single_output_file, args.fail_save, args.paired, args.cutoff, args.force, corrupt_files, report_fp) if args.report_total_kmers: print >> sys.stderr, 'Total number of unique k-mers: {0}'.format( htable.n_unique_kmers()) if args.savetable: print 'Saving k-mer counting table through', input_filename print '...saving to', args.savetable htable.save(args.savetable) fp_rate = \ khmer.calc_expected_collisions(htable, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 print >> sys.stderr, \ 'fp rate estimated to be {fpr:1.3f}'.format(fpr=fp_rate) if args.force and len(corrupt_files) > 0: print >> sys.stderr, "** WARNING: Finished with errors!" print >> sys.stderr, "** IOErrors occurred in the following files:" print >> sys.stderr, "\t", " ".join(corrupt_files)
def main(): parser = sanitize_help(get_parser()) args = parser.parse_args() if not args.quiet: info('trim-low-abund.py', ['streaming']) configure_logging(args.quiet) ### if len(set(args.input_filenames)) != len(args.input_filenames): log_error("Error: Cannot input the same filename multiple times.") sys.exit(1) if args.trim_at_coverage != DEFAULT_TRIM_AT_COVERAGE and \ not args.variable_coverage: log_error("Error: --trim-at-coverage/-Z given, but " "--variable-coverage/-V not specified.") sys.exit(1) if args.diginorm_coverage != DEFAULT_DIGINORM_COVERAGE and \ not args.diginorm: log_error("Error: --diginorm-coverage given, but " "--diginorm not specified.") sys.exit(1) if args.diginorm and args.single_pass: log_error("Error: --diginorm and --single-pass are incompatible!\n" "You probably want to use normalize-by-median.py instead.") sys.exit(1) ### report_on_config(args) check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, args.force) if args.savegraph: graphsize = calculate_graphsize(args, 'countgraph') check_space_for_graph(args.savegraph, graphsize, args.force) if ('-' in args.input_filenames or '/dev/stdin' in args.input_filenames) \ and not args.output: log_error("Accepting input from stdin; output filename must " "be provided with -o.") sys.exit(1) if args.loadgraph: log_info('loading countgraph from {graph}', graph=args.loadgraph) ct = khmer.load_countgraph(args.loadgraph) else: log_info('making countgraph') ct = khmer_args.create_countgraph(args) K = ct.ksize() tempdir = tempfile.mkdtemp('khmer', 'tmp', args.tempdir) log_info( 'created temporary directory {temp};\n' 'use -T to change location', temp=tempdir) trimmer = Trimmer(ct, not args.variable_coverage, args.cutoff, args.trim_at_coverage) if args.diginorm: trimmer.set_diginorm(args.diginorm_coverage) # ### FIRST PASS ### save_pass2_total = 0 written_bp = 0 written_reads = 0 # only create the file writer once if outfp is specified; otherwise, # create it for each file. if args.output: trimfp = get_file_writer(args.output, args.gzip, args.bzip) pass2list = [] for filename in args.input_filenames: # figure out temporary filename for 2nd pass pass2filename = os.path.basename(filename) + '.pass2' pass2filename = os.path.join(tempdir, pass2filename) pass2fp = open(pass2filename, 'w') # construct output filenames if args.output is None: # note: this will be saved in trimfp. outfp = open(os.path.basename(filename) + '.abundtrim', 'wb') # get file handle w/gzip, bzip trimfp = get_file_writer(outfp, args.gzip, args.bzip) # record all this info pass2list.append((filename, pass2filename, trimfp)) # input file stuff: get a broken_paired reader. screed_iter = screed.open(filename) paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=args.ignore_pairs) # main loop through the file. n_start = trimmer.n_reads save_start = trimmer.n_saved watermark = REPORT_EVERY_N_READS for read in trimmer.pass1(paired_iter, pass2fp): if (trimmer.n_reads - n_start) > watermark: log_info( "... {filename} {n_saved} {n_reads} {n_bp} " "{w_reads} {w_bp}", filename=filename, n_saved=trimmer.n_saved, n_reads=trimmer.n_reads, n_bp=trimmer.n_bp, w_reads=written_reads, w_bp=written_bp) watermark += REPORT_EVERY_N_READS # write out the trimmed/etc sequences that AREN'T going to be # revisited in a 2nd pass. write_record(read, trimfp) written_bp += len(read) written_reads += 1 pass2fp.close() log_info("{filename}: kept aside {kept} of {total} from first pass", filename=filename, kept=trimmer.n_saved - save_start, total=trimmer.n_reads - n_start) # first pass goes across all the data, so record relevant stats... n_reads = trimmer.n_reads n_bp = trimmer.n_bp n_skipped = trimmer.n_skipped bp_skipped = trimmer.bp_skipped save_pass2_total = trimmer.n_saved # ### SECOND PASS. ### # nothing should have been skipped yet! assert trimmer.n_skipped == 0 assert trimmer.bp_skipped == 0 if args.single_pass: pass2list = [] # go back through all the files again. for _, pass2filename, trimfp in pass2list: log_info('second pass: looking at sequences kept aside in {pass2}', pass2=pass2filename) # note that for this second pass, we don't care about paired # reads - they will be output in the same order they're read in, # so pairs will stay together if not orphaned. This is in contrast # to the first loop. Hence, force_single=True below. screed_iter = screed.open(pass2filename, parse_description=False) paired_iter = broken_paired_reader(screed_iter, min_length=K, force_single=True) watermark = REPORT_EVERY_N_READS for read in trimmer.pass2(paired_iter): if (trimmer.n_reads - n_start) > watermark: log_info('... x 2 {a} {b} {c} {d} {e} {f} {g}', a=trimmer.n_reads - n_start, b=pass2filename, c=trimmer.n_saved, d=trimmer.n_reads, e=trimmer.n_bp, f=written_reads, g=written_bp) watermark += REPORT_EVERY_N_READS write_record(read, trimfp) written_reads += 1 written_bp += len(read) log_info('removing {pass2}', pass2=pass2filename) os.unlink(pass2filename) # if we created our own trimfps, close 'em. if not args.output: trimfp.close() log_info('removing temp directory & contents ({temp})', temp=tempdir) shutil.rmtree(tempdir) trimmed_reads = trimmer.trimmed_reads n_passes = 1.0 + (float(save_pass2_total) / n_reads) percent_reads_trimmed = float(trimmed_reads + (n_reads - written_reads)) /\ n_reads * 100.0 log_info('read {read} reads, {bp} bp', read=n_reads, bp=n_bp) log_info('wrote {wr} reads, {wbp} bp', wr=written_reads, wbp=written_bp) log_info('looked at {st} reads twice ({np:.2f} passes)', st=save_pass2_total, np=n_passes) log_info('removed {r} reads and trimmed {t} reads ({p:.2f}%)', r=n_reads - written_reads, t=trimmed_reads, p=percent_reads_trimmed) log_info('trimmed or removed {p:.2f}%% of bases ({bp} total)', p=(1 - (written_bp / float(n_bp))) * 100.0, bp=n_bp - written_bp) if args.variable_coverage: percent_reads_hicov = 100.0 * float(n_reads - n_skipped) / n_reads log_info('{n} reads were high coverage ({p:.2f}%);', n=n_reads - n_skipped, p=percent_reads_hicov) log_info('skipped {r} reads/{bp} bases because of low coverage', r=n_skipped, bp=bp_skipped) fp_rate = \ khmer.calc_expected_collisions(ct, args.force, max_false_pos=.8) # for max_false_pos see Zhang et al., http://arxiv.org/abs/1309.2975 log_info('fp rate estimated to be {fpr:1.3f}', fpr=fp_rate) log_info('output in *.abundtrim') if args.savegraph: log_info("Saving k-mer countgraph to {graph}", graph=args.savegraph) ct.save(args.savegraph)
def main(): info('sweep-reads-buffered.py', ['sweep']) parser = get_parser() args = parser.parse_args() if args.min_tablesize < MIN_HSIZE: args.min_tablesize = MIN_HSIZE if args.ksize < MIN_KSIZE: args.ksize = MIN_KSIZE report_on_config(args, hashtype='hashbits') K = args.ksize HT_SIZE = args.min_tablesize N_HT = args.n_tables traversal_range = args.traversal_range input_fastp = args.input_fastp if not args.outdir: outdir = os.path.dirname(input_fastp) else: outdir = args.outdir max_buffers = args.max_buffers output_pref = args.output_prefix buf_size = args.buffer_size max_reads = args.max_reads check_input_files(args.input_fastp, args.force) check_valid_file_exists(args.input_files) all_input_files = [input_fastp] all_input_files.extend(args.input_files) # Check disk space availability check_space(all_input_files, args.force) # figure out input file type (FA/FQ) -- based on first file ix = iter(screed.open(args.input_files[0])) record = ix.next() del ix extension = 'fa' if hasattr(record, 'quality'): # fastq! extension = 'fq' output_buffer = ReadBufferManager( max_buffers, max_reads, buf_size, output_pref, outdir, extension) # consume the partitioned fasta with which to label the graph ht = khmer.LabelHash(K, HT_SIZE, N_HT) try: print >>sys.stderr, 'consuming input sequences...' if args.label_by_pid: print >>sys.stderr, '...labeling by partition id (pid)' ht.consume_partitioned_fasta_and_tag_with_labels(input_fastp) elif args.label_by_seq: print >>sys.stderr, '...labeling by sequence' for n, record in enumerate(screed.open(input_fastp)): if n % 50000 == 0: print >>sys.stderr, \ '...consumed {n} sequences...'.format(n=n) ht.consume_sequence_and_tag_with_labels(record.sequence, n) else: print >>sys.stderr, \ '...labeling to create groups of size {s}'.format( s=args.group_size) label = -1 g = 0 try: outfp = open('{pref}_base_{g}.{ext}'.format(pref=output_pref, g=g, ext=extension ), 'wb') for n, record in enumerate(screed.open(input_fastp)): if n % args.group_size == 0: label += 1 if label > g: g = label outfp = open('{pref}_base_{g}.{ext}'.format( pref=output_pref, g=g, ext=extension), 'wb') if n % 50000 == 0: print >>sys.stderr, \ '...consumed {n} sequences...'.format(n=n) ht.consume_sequence_and_tag_with_labels(record.sequence, label) write_record(record, outfp) except IOError as e: print >>sys.stderr, '!! ERROR !!', e print >>sys.stderr, '...error splitting input. exiting...' except IOError as e: print >>sys.stderr, '!! ERROR: !!', e print >>sys.stderr, '...error consuming \ {i}. exiting...'.format(i=input_fastp) print >>sys.stderr, 'done consuming input sequence. \ added {t} tags and {l} \ labels...'.format(t=ht.n_tags(), l=ht.n_labels()) label_dict = defaultdict(int) label_number_dist = [] n_orphaned = 0 n_labeled = 0 n_mlabeled = 0 total_t = time.clock() start_t = time.clock() for read_file in args.input_files: print >>sys.stderr, '** sweeping {read_file} for labels...'.format( read_file=read_file) file_t = 0.0 try: read_fp = screed.open(read_file) except IOError as error: print >>sys.stderr, '!! ERROR: !!', error print >>sys.stderr, '*** Could not open {fn}, skipping...'.format( fn=read_file) else: for _, record in enumerate(read_fp): if _ % 50000 == 0: end_t = time.clock() batch_t = end_t - start_t file_t += batch_t print >>sys.stderr, '\tswept {n} reads [{nc} labeled, \ {no} orphaned] \ ** {sec}s ({sect}s total)' \ .format(n=_, nc=n_labeled, no=n_orphaned, sec=batch_t, sect=file_t) start_t = time.clock() seq = record.sequence name = record.name try: labels = ht.sweep_label_neighborhood(seq, traversal_range) except ValueError as e: pass else: if hasattr(record, 'quality'): seq_str = fmt_fastq(name, seq, record.quality, labels) else: seq_str = fmt_fasta(name, seq, labels) label_number_dist.append(len(labels)) if labels: n_labeled += 1 if len(labels) > 1: output_buffer.queue(seq_str, 'multi') n_mlabeled += 1 label_dict['multi'] += 1 else: output_buffer.queue(seq_str, labels[0]) label_dict[labels[0]] += 1 else: n_orphaned += 1 output_buffer.queue(seq_str, 'orphaned') label_dict['orphaned'] += 1 print >>sys.stderr, '** End of file {fn}...'.format(fn=read_file) output_buffer.flush_all() read_fp.close() # gotta output anything left in the buffers at the end! print >>sys.stderr, '** End of run...' output_buffer.flush_all() total_t = time.clock() - total_t if output_buffer.num_write_errors > 0 or output_buffer.num_file_errors > 0: print >>sys.stderr, '! WARNING: Sweep finished with errors !' print >>sys.stderr, '** {writee} reads not written'.format( writee=output_buffer.num_write_errors) print >>sys.stderr, '** {filee} errors opening files'.format( filee=output_buffer.num_file_errors) print >>sys.stderr, 'swept {n_reads} for labels...'.format( n_reads=n_labeled + n_orphaned) print >>sys.stderr, '...with {nc} labeled and {no} orphaned'.format( nc=n_labeled, no=n_orphaned) print >>sys.stderr, '...and {nmc} multilabeled'.format(nmc=n_mlabeled) print >>sys.stderr, '** outputting label number distribution...' fn = os.path.join(outdir, '{pref}.dist.txt'.format(pref=output_pref)) with open(fn, 'wb') as outfp: for nc in label_number_dist: outfp.write('{nc}\n'.format(nc=nc)) fn = os.path.join(outdir, '{pref}.counts.csv'.format(pref=output_pref)) print >>sys.stderr, '** outputting label read counts...' with open(fn, 'wb') as outfp: for k in label_dict: outfp.write('{l},{c}\n'.format(l=k, c=label_dict[k]))
def main(): # pylint: disable=too-many-branches,too-many-statements info("saturate-by-median.py", ["diginorm"]) args = get_parser().parse_args() report_on_config(args) report_fp = args.report report_frequency = args.report_frequency check_valid_file_exists(args.input_filenames) check_space(args.input_filenames, False) if args.savetable: check_space_for_hashtable(args, "countgraph", False) # list to save error files along with throwing exceptions if args.force: corrupt_files = [] if args.loadtable: print("loading k-mer counting table from", args.loadtable) htable = khmer.load_counting_hash(args.loadtable) else: print("making countgraph") htable = create_countgraph(args) total = 0 discarded = 0 for index, input_filename in enumerate(args.input_filenames): total_acc = 0 discarded_acc = 0 try: total_acc, discarded_acc = normalize_by_median(input_filename, htable, args, report_fp, report_frequency) except IOError as err: handle_error(err, input_filename) if not args.force: print("** Exiting!", file=sys.stderr) sys.exit(1) else: print("*** Skipping error file, moving on...", file=sys.stderr) corrupt_files.append(input_filename) else: if total_acc == 0 and discarded_acc == 0: print("SKIPPED empty file", input_filename) else: total += total_acc discarded += discarded_acc print( "DONE with {inp}; kept {kept} of {total} or {perc:2}%".format( inp=input_filename, kept=total - discarded, total=total, perc=int(100.0 - discarded / float(total) * 100.0), ) ) if args.savetable: print("Saving k-mer counting table through", input_filename) print("...saving to", args.savetable) htable.save(args.savetable) # re: threshold, see Zhang et al., # http://arxiv.org/abs/1309.2975 fp_rate = khmer.calc_expected_collisions(htable, args.force, max_false_pos=0.8) print("fp rate estimated to be {fpr:1.3f}".format(fpr=fp_rate)) if args.force and len(corrupt_files) > 0: print("** WARNING: Finished with errors!", file=sys.stderr) print("** I/O Errors occurred in the following files:", file=sys.stderr) print("\t", " ".join(corrupt_files), file=sys.stderr)