def update(cfg):
tx = time.time()
# validate input files
input_files = validate_input_files(cfg.input_files, cfg.input_directory,
cfg.input_extension, cfg.quiet)
if len(input_files) == 0:
print_log("ERROR: No valid input files found", cfg.quiet)
return False
# Set db prefixes
db_prefix = {
prefix: cfg.db_prefix + "." + prefix
for prefix in ["ibf", "map", "tax", "gnn"]
}
# Set temporary working folder (current or new output)
tmp_output_folder = cfg.output_db_prefix + "_tmp/" if cfg.output_db_prefix else cfg.db_prefix + "_tmp/"
if not set_tmp_folder(tmp_output_folder): return False
# Load .gnn file
gnn = Gnn(file=db_prefix["gnn"])
# If specialization was set on database
if gnn.specialization:
# if not provided by user, use defition of database
if not cfg.specialization: cfg.specialization = gnn.specialization
print_log("Using --specialization " + cfg.specialization, cfg.quiet)
else:
if cfg.specialization:
# If user defined specialization on update but database has none
print_log(
"ERROR: not possible to update a database with --specialization if it was built without it",
cfg.quiet)
return False
# load bins
bins = Bins(taxsbp_ret=gnn.bins,
use_specialization=True if cfg.specialization else False)
# load seqinfo (file or seqids)
seqinfo = load_seqinfo(cfg, input_files)
# check sequences compared to bins
added_seqids, removed_seqids, kept_seqids = check_updated_seqids(
set(seqinfo.get_seqids()), set(bins.get_seqids()))
# Ignore removed sequences if not doing complete update
if cfg.update_complete:
print_log(
"Update: adding " + str(len(added_seqids)) +
" sequences, removing " + str(len(removed_seqids)) +
" sequences, keeping " + str(len(kept_seqids)) + " sequences",
cfg.quiet)
else:
removed_seqids = []
print_log(
"Update: adding " + str(len(added_seqids)) +
" sequences, ignoring " + str(len(kept_seqids)) +
" repeated sequences", cfg.quiet)
print_log("", cfg.quiet)
if not added_seqids and not removed_seqids:
print_log("Nothing to update", cfg.quiet)
rm_tmp_folder(tmp_output_folder)
return False
if cfg.update_complete:
# Remove already included seqids to just retrieve information for added sequences
seqinfo.remove_seqids(kept_seqids | removed_seqids)
else:
# Remove seqids already present in the current version (repeated entries)
seqinfo.remove_seqids(kept_seqids)
# retrive sequence information (after removing invalid seqids)
if not cfg.seq_info_file:
retrieve_seqinfo(seqinfo, tmp_output_folder, input_files, cfg)
# Convert cols data types
if cfg.specialization:
replaced_spec = seqinfo.validate_specialization()
if replaced_spec:
print_log(
str(replaced_spec) +
" invalid specialization entries (sequence accession used instead)\n",
cfg.quiet)
if not cfg.seq_info_file and cfg.write_seq_info_file:
seqinfo.write(cfg.output_db_prefix + ".seqinfo.txt")
# save set of current binids
previous_binids = set(bins.get_binids())
# remove seqids from bins if performing update complete
if cfg.update_complete and removed_seqids:
bins.remove_seqids(removed_seqids)
# save set of kept binids after removal
kept_binids = set(bins.get_binids())
# Set up taxonomy files
ncbi_nodes_file, ncbi_merged_file, ncbi_names_file = set_taxdump_files(
cfg.taxdump_file, tmp_output_folder, cfg.quiet)
tx = time.time()
print_log("Running taxonomic clustering (TaxSBP)", cfg.quiet)
updated_bins = run_taxsbp(seqinfo,
gnn.bin_length,
gnn.fragment_length,
gnn.overlap_length,
gnn.rank,
cfg.specialization,
ncbi_nodes_file,
ncbi_merged_file,
cfg.verbose,
bins=bins)
# bin statistics
taxsbp_binids = set(updated_bins.get_binids())
removed_binids = previous_binids.difference(kept_binids | taxsbp_binids)
new_binids = taxsbp_binids.difference(previous_binids)
updated_binids = taxsbp_binids.intersection(previous_binids)
print_log(
" - " + str(len(new_binids)) + " bins added, " +
str(len(updated_binids)) + " bins updated, " +
str(len(removed_binids)) + " bins removed", cfg.quiet)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
tx = time.time()
print_log("Updating database files", cfg.quiet)
# load new taxonomy
print_log(
" - " + cfg.output_db_prefix +
".tax" if cfg.output_db_prefix else db_prefix["tax"], cfg.quiet)
tax = Tax(ncbi_nodes=ncbi_nodes_file, ncbi_names=ncbi_names_file)
# Update and write .tax file
# filter only used taxids
tax.filter(updated_bins.get_taxids())
# add specialization nodes
if cfg.specialization:
tax.add_nodes(updated_bins.get_specialization_taxid(),
cfg.specialization)
# Load old .tax file into new taxonomy
tax.merge(Tax([db_prefix["tax"]]))
# Write .tax file
tax.write(cfg.output_db_prefix +
".tax" if cfg.output_db_prefix else db_prefix["tax"])
# TODO - remove entries from .tax from removed entries of the db
# merge updated and old bins together
bins.merge(updated_bins)
# Write .gnn file
print_log(
" - " + cfg.output_db_prefix +
".gnn" if cfg.output_db_prefix else db_prefix["gnn"], cfg.quiet)
gnn.bins = bins.get_list() # save updated bins
gnn.number_of_bins = bins.get_number_of_bins() # add new bins count
# set new specialization to gnn
gnn.specialization = cfg.specialization
gnn.write(cfg.output_db_prefix +
".gnn" if cfg.output_db_prefix else db_prefix["gnn"])
# Recreate .map file based on the new bins
print_log(
" - " + cfg.output_db_prefix +
".map" if cfg.output_db_prefix else db_prefix["map"], cfg.quiet)
bins.write_map_file(
cfg.output_db_prefix +
".map" if cfg.output_db_prefix else db_prefix["map"],
use_specialization=True if cfg.specialization else False)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
tx = time.time()
print_log("Updating index (ganon-build)", cfg.quiet)
# Write aux. file for ganon
# This file has to contain all new sequences
# in case of update_complete
acc_bin_file = tmp_output_folder + "acc_bin.txt"
if cfg.update_complete:
# all sequences from the bins with added/removed sequences should be written
bins.write_acc_bin_file(acc_bin_file, new_binids | updated_binids)
# If all sequences of a bin were removed and no new sequence added
# insert a dummy entry for ganon-build to clear the bin
if removed_binids:
with open(acc_bin_file, "a") as abf:
for b in removed_binids:
print(0, 0, 0, b, sep="\t", file=abf)
else:
# Only new sequences (updated_bins) either on old or new binids
updated_bins.write_acc_bin_file(acc_bin_file)
# Update with same values used for build
kmer_size = gnn.kmer_size
window_size = gnn.window_size
hash_functions = gnn.hash_functions
# Free memory for build
del seqinfo
del bins
del updated_bins
del tax
del gnn
# Temporary output filter
tmp_db_prefix_ibf = tmp_output_folder + "ganon.ibf"
run_ganon_build_cmd = " ".join([
cfg.path_exec['build'], "--update-filter-file " + db_prefix["ibf"],
"--kmer-size " + str(kmer_size),
"--window-size " + str(window_size) if window_size else "",
"--count-hashes " if window_size else "", "--hash-functions " +
str(hash_functions), "--seqid-bin-file " + acc_bin_file,
"--output-filter-file " + tmp_db_prefix_ibf,
"--threads " + str(cfg.threads), "--verbose" if cfg.verbose else "",
"--quiet" if cfg.quiet else "", "--n-refs " +
str(cfg.n_refs) if cfg.n_refs is not None else "", "--n-batches " +
str(cfg.n_batches) if cfg.n_batches is not None else "",
"--reference-files " + ",".join([file for file in input_files])
if input_files and not cfg.input_directory else "",
"--directory-reference-files " +
cfg.input_directory if cfg.input_directory else "",
"--extension " + cfg.input_extension if cfg.input_extension else "",
"--update-complete" if cfg.update_complete else ""
])
stdout, stderr = run(run_ganon_build_cmd, print_stderr=True)
# move IBF to final location
shutil.move(
tmp_db_prefix_ibf, cfg.output_db_prefix +
".ibf" if cfg.output_db_prefix else db_prefix["ibf"])
# Delete temp files
rm_tmp_folder(tmp_output_folder)
return True
def update(cfg):
tx = time.time()
# validate input files
input_files = validate_input_files(cfg.input_files, cfg.input_directory,
cfg.input_extension, cfg.quiet)
if len(input_files) == 0:
print_log("ERROR: No valid input files found")
return False
# Set db prefixes
db_prefix = {
prefix: cfg.db_prefix + "." + prefix
for prefix in ["ibf", "map", "tax", "gnn"]
}
# Set temporary working folder (current or new output)
tmp_output_folder = cfg.output_db_prefix + "_tmp/" if cfg.output_db_prefix else cfg.db_prefix + "_tmp/"
if not set_tmp_folder(tmp_output_folder): return False
# Load .gnn file
gnn = Gnn(file=db_prefix["gnn"])
# Set assembly mode
use_assembly = True if gnn.rank == "assembly" else False
# load bins
bins = Bins(taxsbp_ret=gnn.bins)
# Load seqids and generate seqinfo
if cfg.seq_info_file:
seqinfo = load_seqids(seq_info_file=cfg.seq_info_file, quiet=cfg.quiet)
else:
seqinfo = load_seqids(files=input_files, quiet=cfg.quiet)
# check sequences compared to bins
added_seqids, removed_seqids, kept_seqids = check_updated_seqids(
set(seqinfo.get_seqids()), set(bins.get_seqids()))
# Ignore removed sequences if not doing complete update
if cfg.update_complete:
print_log(
"Update: adding " + str(len(added_seqids)) +
" sequences, removing " + str(len(removed_seqids)) +
" sequences, keeping " + str(len(kept_seqids)) + " sequences",
cfg.quiet)
else:
removed_seqids = []
print_log(
"Update: adding " + str(len(added_seqids)) +
" sequences, ignoring " + str(len(kept_seqids)) +
" repeated sequences", cfg.quiet)
print_log("", cfg.quiet)
if not added_seqids and not removed_seqids:
print_log("ERROR: Nothing to update")
return False
if cfg.update_complete:
# Remove already included seqids to just retrieve information for added sequences
seqinfo.remove_seqids(kept_seqids | removed_seqids)
else:
# Remove seqids already present in the current version (repeated entries)
seqinfo.remove_seqids(kept_seqids)
# load seqinfo file with data (after removing ids)
if not cfg.seq_info_file:
load_seqinfo(tmp_output_folder, seqinfo, cfg.path_exec,
cfg.seq_info_mode, use_assembly, cfg.quiet)
if cfg.write_seq_info_file:
seqinfo.write(cfg.output_db_prefix + ".seqinfo.txt" if cfg.
output_db_prefix else cfg.db_prefix + ".seqinfo.txt")
# save set of current binids
previous_binids = set(bins.get_binids())
# remove seqids from bins if performing update complete
if cfg.update_complete and removed_seqids:
bins.remove_seqids(removed_seqids)
# save set of kept binids after removal
kept_binids = set(bins.get_binids())
# Set up taxonomy files
ncbi_nodes_file, ncbi_merged_file, ncbi_names_file = set_taxdump_files(
cfg.taxdump_file, tmp_output_folder, cfg.quiet)
tx = time.time()
print_log("Running taxonomic clustering (TaxSBP)", cfg.quiet)
taxsbp_params = {}
taxsbp_params["update_table"] = bins.get_csv()
taxsbp_params["nodes_file"] = ncbi_nodes_file
taxsbp_params["bin_len"] = gnn.bin_length
if use_assembly:
taxsbp_params["bin_exclusive"] = "assembly"
elif gnn.rank == "taxid":
taxsbp_params["bin_exclusive"] = "leaves"
else:
taxsbp_params["bin_exclusive"] = gnn.rank
if ncbi_merged_file: taxsbp_params["merged_file"] = ncbi_merged_file
if gnn.fragment_length:
taxsbp_params["fragment_len"] = gnn.fragment_length
taxsbp_params["overlap_len"] = gnn.overlap_length
if use_assembly: taxsbp_params["specialization"] = "assembly"
taxsbp_params["input_table"] = seqinfo.get_csv()
updated_bins = Bins(taxsbp_ret=taxsbp.taxsbp.pack(**taxsbp_params))
# bin statistics
taxsbp_binids = set(updated_bins.get_binids())
removed_binids = previous_binids.difference(kept_binids | taxsbp_binids)
new_binids = taxsbp_binids.difference(previous_binids)
updated_binids = taxsbp_binids.intersection(previous_binids)
print_log(
" - " + str(len(new_binids)) + " bins added, " +
str(len(updated_binids)) + " bins updated, " +
str(len(removed_binids)) + " bins removed", cfg.quiet)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
tx = time.time()
print_log("Updating database files", cfg.quiet)
# load new taxonomy
print_log(
" - " + cfg.output_db_prefix +
".tax" if cfg.output_db_prefix else db_prefix["tax"], cfg.quiet)
tax = Tax(ncbi_nodes=ncbi_nodes_file, ncbi_names=ncbi_names_file)
# Update and write .tax file
tax.filter(updated_bins.get_taxids()) # filter only used taxids
if use_assembly:
tax.add_nodes(updated_bins.get_specialization_taxid(),
"assembly") # add assembly nodes
# Load old .tax file into new taxonomy
tax.merge(Tax([db_prefix["tax"]]))
# Write .tax file
tax.write(cfg.output_db_prefix +
".tax" if cfg.output_db_prefix else db_prefix["tax"])
# TODO - remove entries from .tax from removed entries of the db
# merge updated and old bins together
bins.merge(updated_bins)
# Write .gnn file
print_log(
" - " + cfg.output_db_prefix +
".gnn" if cfg.output_db_prefix else db_prefix["gnn"], cfg.quiet)
gnn.bins = bins.get_list() # save updated bins
gnn.number_of_bins = bins.get_number_of_bins() # add new bins count
gnn.write(cfg.output_db_prefix +
".gnn" if cfg.output_db_prefix else db_prefix["gnn"])
# Recreate .map file based on the new bins
print_log(
" - " + cfg.output_db_prefix +
".map" if cfg.output_db_prefix else db_prefix["map"], cfg.quiet)
bins.write_map_file(
cfg.output_db_prefix +
".map" if cfg.output_db_prefix else db_prefix["map"], use_assembly)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
tx = time.time()
print_log("Updating index (ganon-build)", cfg.quiet)
# Write aux. file for ganon
# This file has to contain all new sequences
# in case of update_complete,
acc_bin_file = tmp_output_folder + "acc_bin.txt"
if cfg.update_complete:
# all sequences from the bins with added/removed sequences should be written
bins.write_acc_bin_file(acc_bin_file, new_binids | updated_binids)
# If all sequences of a bin were removed and no new sequence added
# insert a dummy entry for ganon-build to clear the bin
if removed_binids:
with open(acc_bin_file, "a") as abf:
for b in removed_binids:
print(0, 0, 0, b, sep="\t", file=abf)
else:
# Only new sequences (updated_bins) either on old or new binids
updated_bins.write_acc_bin_file(acc_bin_file)
# Temporary output filter
tmp_db_prefix_ibf = tmp_output_folder + "ganon.ibf"
run_ganon_build_cmd = " ".join([
cfg.path_exec['build'], "--update-filter-file " + db_prefix["ibf"],
"--seqid-bin-file " + acc_bin_file, "--output-filter-file " +
tmp_db_prefix_ibf, "--threads " + str(cfg.threads),
"--verbose" if cfg.verbose else "", "--quiet" if cfg.quiet else "",
"--n-refs " + str(cfg.n_refs) if cfg.n_refs is not None else "",
"--n-batches " +
str(cfg.n_batches) if cfg.n_batches is not None else "",
"--reference-files " +
",".join([file for file in input_files]) if input_files else "",
"--directory-reference-files " +
cfg.input_directory if cfg.input_directory else "",
"--extension " + cfg.input_extension if cfg.input_extension else "",
"--update-complete" if cfg.update_complete else ""
])
stdout, stderr = run(run_ganon_build_cmd, print_stderr=True)
# move IBF to final location
shutil.move(
tmp_db_prefix_ibf, cfg.output_db_prefix +
".ibf" if cfg.output_db_prefix else db_prefix["ibf"])
# Delete temp files
rm_tmp_folder(tmp_output_folder)
return True
def build(cfg):
# validate input files
input_files = validate_input_files(cfg.input_files, cfg.input_directory,
cfg.input_extension, cfg.quiet)
if len(input_files) == 0:
print_log("ERROR: No valid input files found", cfg.quiet)
return False
# Set db prefixes
db_prefix = {
prefix: cfg.db_prefix + "." + prefix
for prefix in ["ibf", "map", "tax", "gnn"]
}
# Set temporary working folder
tmp_output_folder = cfg.db_prefix + "_tmp/"
if not set_tmp_folder(tmp_output_folder): return False
# Set up taxonomy
ncbi_nodes_file, ncbi_merged_file, ncbi_names_file = set_taxdump_files(
cfg.taxdump_file, tmp_output_folder, cfg.quiet)
# Parse .tax
tx = time.time()
print_log("Parsing taxonomy", cfg.quiet)
tax = Tax(ncbi_nodes=ncbi_nodes_file, ncbi_names=ncbi_names_file)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
# load seqinfo (file or seqids)
seqinfo = load_seqinfo(cfg, input_files)
# Retrieve sequence information
if not cfg.seq_info_file:
retrieve_seqinfo(seqinfo, tmp_output_folder, input_files, cfg)
# Check for valid specialization
if cfg.specialization:
replaced_spec = seqinfo.validate_specialization()
if replaced_spec:
print_log(
str(replaced_spec) +
" invalid specialization entries (sequence accession used instead)\n",
cfg.quiet)
# Write seq-info-file
if not cfg.seq_info_file and cfg.write_seq_info_file:
seqinfo.write(cfg.db_prefix + ".seqinfo.txt")
# check sequences compared to bins
added_seqids, _, _ = check_updated_seqids(set(seqinfo.get_seqids()), set())
# Ignore removed sequences if not doing complete update
print_log("Build: adding " + str(len(added_seqids)) + " sequences",
cfg.quiet)
print_log("", cfg.quiet)
if not added_seqids:
print_log("No valid seq. info to build", cfg.quiet)
rm_tmp_folder(tmp_output_folder)
return False
# Set or calculate best --bin-length
if cfg.bin_length:
bin_length = cfg.bin_length
else:
tx = time.time()
print_log("Simulating parameters", cfg.quiet)
bin_length = estimate_bin_length(cfg, seqinfo, tax)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
# Set fragment length
if cfg.fragment_length == -1: # if ==-1 set default
fragment_length = bin_length - cfg.overlap_length
elif cfg.fragment_length == 0: # if ==0 deactivate
fragment_length = 0
else: # user input
fragment_length = cfg.fragment_length - cfg.overlap_length
tx = time.time()
print_log("Running taxonomic clustering (TaxSBP)", cfg.quiet)
bins = run_taxsbp(seqinfo, bin_length, fragment_length, cfg.overlap_length,
cfg.rank, cfg.specialization, ncbi_nodes_file,
ncbi_merged_file, cfg.verbose)
# bin statistics
actual_number_of_bins = bins.get_number_of_bins()
optimal_number_of_bins = optimal_bins(actual_number_of_bins)
max_length_bin = bins.get_max_bin_length()
max_kmer_count = estimate_elements(max_length_bin, cfg.kmer_size,
cfg.window_size)
#max_kmer_count = max_length_bin - cfg.kmer_size + 1
print_log(" - " + str(actual_number_of_bins) + " bins created", cfg.quiet)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
# Get optimal parameters from user input and taxsbp result
if cfg.filter_size:
optimal_params = derive_bf_params(
max_kmer_count, 0,
math.ceil(mb2bits(cfg.filter_size) / optimal_number_of_bins),
cfg.hash_functions)
else:
optimal_params = derive_bf_params(max_kmer_count, cfg.max_fp, 0,
cfg.hash_functions)
# When fixed size is too small
if optimal_params["hash_functions"] == 0:
optimal_params["hash_functions"] = 3
print_log("Optimal bins: " + str(optimal_number_of_bins), cfg.quiet)
print_log(
"Max. false positive: " +
str("{0:.5f}".format(optimal_params["false_positive"])), cfg.quiet)
print_log("Hash functions: " + str(optimal_params["hash_functions"]),
cfg.quiet)
if cfg.window_size:
# Check lower bound for minimizers with estimated bin-length
min_mini = math.floor(((max_length_bin - cfg.kmer_size + 1) /
(cfg.window_size - cfg.kmer_size + 1)))
min_size_mini = derive_bf_params(min_mini,
optimal_params["false_positive"], 0,
optimal_params["hash_functions"])
print_log(
"Possible elements per bin: " + str(min_mini) + ".." +
str(max_kmer_count), cfg.quiet)
print_log(
"Possible filter sizes: " + str("{0:.2f}".format(
bits2mb(min_size_mini["size_bits"] * optimal_number_of_bins)))
+ "MB.." + str("{0:.2f}".format(
bits2mb(optimal_params["size_bits"] * optimal_number_of_bins)))
+ "MB", cfg.quiet)
else:
print_log("Elements per bin: " + str(max_kmer_count), cfg.quiet)
print_log(
"Filter size: " + str("{0:.2f}".format(
bits2mb(optimal_params["size_bits"] * optimal_number_of_bins)))
+ "MB", cfg.quiet)
print_log("")
# Build database files (map, tax, gnn)
tx = time.time()
print_log("Building database files", cfg.quiet)
# Write .map file
print_log(" - " + db_prefix["map"], cfg.quiet)
bins.write_map_file(
db_prefix["map"],
use_specialization=True if cfg.specialization else False)
# Write .tax file
print_log(" - " + db_prefix["tax"], cfg.quiet)
# filter only used taxids
tax.filter(bins.get_taxids())
# add specialization nodes
if cfg.specialization:
tax.add_nodes(bins.get_specialization_taxid(), cfg.specialization)
tax.write(db_prefix["tax"])
if cfg.specialization and cfg.rank != "leaves":
print_log(
" - --rank is set to leaves when using specialization values",
cfg.quiet)
cfg.rank = "leaves"
# Write .gnn file
print_log(" - " + db_prefix["gnn"], cfg.quiet)
gnn = Gnn(kmer_size=cfg.kmer_size,
window_size=cfg.window_size,
hash_functions=optimal_params["hash_functions"],
number_of_bins=actual_number_of_bins,
rank=cfg.rank,
specialization=cfg.specialization,
bin_length=bin_length,
fragment_length=fragment_length,
overlap_length=cfg.overlap_length,
bins=bins.get_list())
gnn.write(db_prefix["gnn"])
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
print_log("Building index (ganon-build)", cfg.quiet)
# Write aux. file for ganon
acc_bin_file = tmp_output_folder + "acc_bin.txt"
bins.write_acc_bin_file(acc_bin_file)
# Free memory for build
del seqinfo
del bins
del tax
del gnn
run_ganon_build_cmd = " ".join([
cfg.path_exec['build'], "--seqid-bin-file " + acc_bin_file,
"--bin-size-bits " + str(optimal_params["size_bits"])
if cfg.filter_size else "--false-positive " + str(cfg.max_fp),
"--kmer-size " + str(cfg.kmer_size),
"--window-size " + str(cfg.window_size) if cfg.window_size else "",
"--count-hashes " if cfg.window_size else "", "--hash-functions " +
str(optimal_params["hash_functions"]), "--threads " + str(cfg.threads),
"--output-filter-file " + db_prefix["ibf"],
"--verbose" if cfg.verbose else "", "--quiet" if cfg.quiet else "",
"--n-refs " + str(cfg.n_refs) if cfg.n_refs is not None else "",
"--n-batches " +
str(cfg.n_batches) if cfg.n_batches is not None else "",
"--reference-files " + ",".join([file for file in input_files])
if input_files and not cfg.input_directory else "",
"--directory-reference-files " +
cfg.input_directory if cfg.input_directory else "",
"--extension " + cfg.input_extension if cfg.input_extension else ""
])
stdout, stderr = run(run_ganon_build_cmd, print_stderr=True)
# Delete temp files
rm_tmp_folder(tmp_output_folder)
return True
def build(cfg):
# validate input files
input_files = validate_input_files(cfg.input_files, cfg.input_directory,
cfg.input_extension, cfg.quiet)
if len(input_files) == 0:
print_log("ERROR: No valid input files found")
return False
# Set db prefixes
db_prefix = {
prefix: cfg.db_prefix + "." + prefix
for prefix in ["ibf", "map", "tax", "gnn"]
}
# Set temporary working folder
tmp_output_folder = cfg.db_prefix + "_tmp/"
if not set_tmp_folder(tmp_output_folder): return False
# Set assembly mode
use_assembly = True if cfg.rank == "assembly" else False
# Set up taxonomy
ncbi_nodes_file, ncbi_merged_file, ncbi_names_file = set_taxdump_files(
cfg.taxdump_file, tmp_output_folder, cfg.quiet)
tx = time.time()
print_log("Parsing taxonomy", cfg.quiet)
tax = Tax(ncbi_nodes=ncbi_nodes_file, ncbi_names=ncbi_names_file)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
# Load seqids and generate seqinfo
if cfg.seq_info_file:
seqinfo = load_seqids(seq_info_file=cfg.seq_info_file, quiet=cfg.quiet)
else:
seqinfo = load_seqids(files=input_files, quiet=cfg.quiet)
load_seqinfo(tmp_output_folder, seqinfo, cfg.path_exec,
cfg.seq_info_mode, use_assembly, cfg.quiet)
if cfg.write_seq_info_file:
seqinfo.write(cfg.db_prefix + ".seqinfo.txt")
# check sequences compared to bins
added_seqids, _, _ = check_updated_seqids(set(seqinfo.get_seqids()), set())
# Ignore removed sequences if not doing complete update
print_log("Build: adding " + str(len(added_seqids)) + " sequences",
cfg.quiet)
print_log("", cfg.quiet)
# Set bin length
if cfg.bin_length: # user defined
bin_length = cfg.bin_length
else:
tx = time.time()
print_log("Calculating best bin length", cfg.quiet)
bin_length, approx_size, n_bins = estimate_bin_len_size(
cfg, seqinfo, tax, use_assembly)
if bin_length <= 0:
bin_length = 1000000
print_log(
"WARNING: could not estimate bin length, using default of " +
str(bin_length) + "bp")
else:
print_log(
" - bin length: " + str(bin_length) + "bp (approx: " +
str(n_bins) + " bins / " + str("{0:.2f}".format(approx_size)) +
"MB)", cfg.quiet)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
# Set fragment length
if cfg.fragment_length == -1: # if ==-1 set default
fragment_length = bin_length - cfg.overlap_length
elif cfg.fragment_length == 0: # if ==0 deactivate
fragment_length = 0
else: # user input
fragment_length = cfg.fragment_length - cfg.overlap_length
tx = time.time()
print_log("Running taxonomic clustering (TaxSBP)", cfg.quiet)
taxsbp_params = {}
taxsbp_params["nodes_file"] = ncbi_nodes_file
taxsbp_params["bin_len"] = bin_length
if use_assembly:
taxsbp_params["bin_exclusive"] = "assembly"
elif cfg.rank == "taxid":
taxsbp_params["bin_exclusive"] = "leaves"
else:
taxsbp_params["bin_exclusive"] = cfg.rank
if ncbi_merged_file: taxsbp_params["merged_file"] = ncbi_merged_file
if fragment_length:
taxsbp_params["fragment_len"] = fragment_length
taxsbp_params["overlap_len"] = cfg.overlap_length
if use_assembly: taxsbp_params["specialization"] = "assembly"
taxsbp_params["input_table"] = seqinfo.get_csv()
bins = Bins(taxsbp_ret=taxsbp.taxsbp.pack(**taxsbp_params))
del taxsbp_params
# bin statistics
actual_number_of_bins = bins.get_number_of_bins()
optimal_number_of_bins = optimal_bins(actual_number_of_bins)
max_length_bin = bins.get_max_bin_length()
max_kmer_count = max_length_bin - cfg.kmer_size + 1 # aproximate number of unique k-mers by just considering that they are all unique
print_log(" - " + str(actual_number_of_bins) + " bins created", cfg.quiet)
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
tx = time.time()
print_log("Building database files", cfg.quiet)
# Write .map file
print_log(" - " + db_prefix["map"], cfg.quiet)
bins.write_map_file(db_prefix["map"], use_assembly)
# Write .tax file
print_log(" - " + db_prefix["tax"], cfg.quiet)
tax.filter(bins.get_taxids()) # filter only used taxids
if use_assembly:
tax.add_nodes(bins.get_specialization_taxid(),
"assembly") # add assembly nodes
tax.write(db_prefix["tax"])
# Write .gnn file
print_log(" - " + db_prefix["gnn"], cfg.quiet)
gnn = Gnn(kmer_size=cfg.kmer_size,
hash_functions=cfg.hash_functions,
number_of_bins=actual_number_of_bins,
rank=cfg.rank,
bin_length=bin_length,
fragment_length=fragment_length,
overlap_length=cfg.overlap_length,
bins=bins.get_list())
gnn.write(db_prefix["gnn"])
print_log(" - done in " + str("%.2f" % (time.time() - tx)) + "s.\n",
cfg.quiet)
print_log("Building index (ganon-build)", cfg.quiet)
# define bloom filter size based on given false positive
MBinBits = 8388608
print_log(
" - max unique " + str(cfg.kmer_size) + "-mers: " +
str(max_kmer_count), cfg.quiet)
if not cfg.fixed_bloom_size:
bin_size_bits = math.ceil(-(1 / (
(1 - cfg.max_fp**(1 / float(cfg.hash_functions)))**
(1 / float(cfg.hash_functions * max_kmer_count)) - 1)))
print_log(
" - IBF calculated size with fp<=" + str(cfg.max_fp) + ": " +
str("{0:.2f}".format(
(bin_size_bits * optimal_number_of_bins) / MBinBits)) +
"MB (" + str(bin_size_bits) + " bits/bin * " +
str(optimal_number_of_bins) + " optimal bins [" +
str(actual_number_of_bins) + " real bins])", cfg.quiet)
else:
bin_size_bits = math.ceil(
(cfg.fixed_bloom_size * MBinBits) / optimal_number_of_bins)
estimated_max_fp = (1 - ((1 - (1 / float(bin_size_bits)))**(
cfg.hash_functions * max_kmer_count)))**cfg.hash_functions
print_log(
" - IBF calculated max. fp with size=" +
str(cfg.fixed_bloom_size) + "MB: " +
str("{0:.2f}".format(estimated_max_fp) + " (" +
str(optimal_number_of_bins) + " optimal bins [" +
str(actual_number_of_bins) + " real bins])"), cfg.quiet)
# Write aux. file for ganon
acc_bin_file = tmp_output_folder + "acc_bin.txt"
bins.write_acc_bin_file(acc_bin_file)
run_ganon_build_cmd = " ".join([
cfg.path_exec['build'], "--seqid-bin-file " + acc_bin_file,
"--filter-size-bits " + str(bin_size_bits * optimal_number_of_bins)
if cfg.max_fp else "--filter-size " + str(cfg.fixed_bloom_size),
"--kmer-size " + str(cfg.kmer_size), "--hash-functions " +
str(cfg.hash_functions), "--threads " + str(cfg.threads),
"--output-filter-file " + db_prefix["ibf"],
"--verbose" if cfg.verbose else "", "--quiet" if cfg.quiet else "",
"--n-refs " + str(cfg.n_refs) if cfg.n_refs is not None else "",
"--n-batches " +
str(cfg.n_batches) if cfg.n_batches is not None else "",
"--reference-files " +
",".join([file for file in input_files]) if input_files else "",
"--directory-reference-files " +
cfg.input_directory if cfg.input_directory else "",
"--extension " + cfg.input_extension if cfg.input_extension else ""
])
stdout, stderr = run(run_ganon_build_cmd, print_stderr=True)
# Delete temp files
rm_tmp_folder(tmp_output_folder)
return True