def run_epa_ng(tree: str,
ref_msa_fastafile: str,
study_msa_fastafile: str,
model: str,
out_dir: str,
chunk_size=5000,
threads=1,
print_cmds=False):
'''Run EPA-NG on specified tree, reference MSA, and study sequence MSA.
Will output a .jplace file in out_dir.'''
make_output_dir(out_dir)
epa_ng_command = [
"epa-ng", "--tree", tree, "--ref-msa", ref_msa_fastafile, "--query",
study_msa_fastafile, "--chunk-size",
str(chunk_size), "-T",
str(threads), "-m", model, "-w", out_dir, "--filter-acc-lwr", "0.99",
"--filter-max", "100"
]
system_call_check(epa_ng_command, print_out=print_cmds)
# Parse jplace file so that output is reprodicible.
jplace_orig = path.join(out_dir, "epa_result.jplace")
jplace_parsed = path.join(out_dir, "epa_result_parsed.jplace")
parse_jplace(jplace_orig, jplace_parsed)
def castor_nsti(tree_path,
known_tips):
'''Will calculate distance from each study sequence to the closest
reference sequence. Takes in the path to treefile and the known tips
(i.e. the rownames in the trait table - the reference genome ids).'''
castor_nsti_script = path.join(path.dirname(path.abspath(__file__)),
'Rscripts', 'castor_nsti.R')
# Create temporary directory for working in.
with TemporaryDirectory() as temp_dir:
# Output known tip names to temp file
# (note this object is a numpy.ndarray)
known_tips_out = path.join(temp_dir, "known_tips.txt")
known_tips.tofile(known_tips_out, sep="\n")
nsti_tmp_out = path.join(temp_dir, "nsti_out.txt")
# Run Rscript.
system_call_check(" ".join(["Rscript",
castor_nsti_script,
tree_path,
known_tips_out,
nsti_tmp_out]))
# Read in calculated NSTI values.
nsti_out = pd.read_csv(nsti_tmp_out, sep="\t", index_col="sequence")
# Make sure that the table has the correct number of rows.
if len(known_tips) != nsti_out.shape[0]:
ValueError("Number of rows in returned NSTI table is incorrect.")
return(nsti_out)
def run_papara(tree: str, ref_msa: dict, study_fasta: str, out_dir: str,
threads=1, print_cmds=False):
'''Run PaPaRa to place study sequences into reference multiple-sequence
alignment (MSA). Will return dictionary of the the output MSA (sequence ids
as keys). Expects path to tree and study FASTA as strings. Expects
reference MSA as a dictionary output by read_fasta. This MSA will be
converted to phylip format before running PaPaRa.'''
# Get absolute paths to input files.
tree = path.abspath(tree)
study_fasta = path.abspath(study_fasta)
# Change working directory to out directory (but keep track of original).
# This is necessary because PaPaRa outputs into the current working
# directory.
orig_wd = getcwd()
chdir(out_dir)
# Convert ref sequences from MSA FASTA to phylip.
write_phylip(ref_msa, "ref_seqs.phylip")
# Make call to papara to place sequences (outputs phylip format).
system_call_check("papara -t " + tree + " -s ref_seqs.phylip " +
"-q " + study_fasta + " -j " + str(threads) +
" -n out", print_command=print_cmds,
print_stdout=print_cmds, print_stderr=print_cmds)
# Change back to original working directory.
chdir(orig_wd)
# Read in papara phylip output and return.
return(read_phylip(path.join(out_dir, "papara_alignment.out"),
check_input=True))
def test_full_pipeline_tsv(self):
'''Test that full pipeline can be run without error with
TSV sequence abundance table.'''
with TemporaryDirectory() as temp_dir:
out_tree = path.join(temp_dir, "out.tre")
system_call_check("place_seqs.py -s " + test_study_seqs + " -r " +
test_ref_dir + " -o " + out_tree)
traits_predict = path.join(temp_dir, "hsp_out.tsv.gz")
marker_predict = path.join(temp_dir, "hsp_out_marker.tsv.gz")
system_call_check("hsp.py -t " + out_tree +
" --observed_trait_table " + test_known_traits +
" -n -o " + traits_predict)
system_call_check("hsp.py -t " + out_tree +
" --observed_trait_table " + test_known_marker +
" -n -o " + marker_predict)
metagenome_out = path.join(temp_dir, "meta_out")
system_call_check("metagenome_pipeline.py -i " +
test_seq_abun_tsv + " -f " + traits_predict +
" --strat_out " + " -m " + marker_predict +
" -o " + metagenome_out)
metagenome_outfile = path.join(metagenome_out,
"pred_metagenome_unstrat.tsv.gz")
system_call_check("pathway_pipeline.py -i " + metagenome_outfile +
" -o " + temp_dir)
def test_picrust2_pipeline_script_per_seq_contrib_strat(self):
'''Test that full pipeline can be run successfully with
picrust2_pipeline.py with the --per_sequence_contrib and --stratified
options.'''
with TemporaryDirectory() as temp_dir:
out_dir = path.join(temp_dir, "pipeline_out")
system_call_check("picrust2_pipeline.py -s " + test_study_seqs +
" -i " + test_seq_abun_tsv +
" -o " + out_dir +
" -r " + test_ref_dir +
" -p 1" +
" --custom_trait_tables " + test_known_traits +
" --marker_gene_table " + test_known_marker +
" --reaction_func " + test_known_traits +
" --max_nsti 1.9" +
" --min_reads 2" +
" --min_samples 2" +
" --skip_minpath" +
" --no_gap_fill" +
" --coverage" +
" --remove_intermediate" +
" --stratified" +
" --per_sequence_contrib"
" --verbose")
def minpath_wrapper(sample_id,
unstrat_input,
minpath_map,
minpath_outdir,
print_opt=False,
extra_str=""):
'''Run MinPath based on gene abundances in a single sample. Will return
a set of all pathways called as present.'''
# Define MinPath input and output filenames.
minpath_in = path.join(minpath_outdir,
str(sample_id) + extra_str + "_minpath_in.txt")
minpath_report = path.join(
minpath_outdir,
str(sample_id) + extra_str + "_minpath_report.txt")
minpath_details = path.join(
minpath_outdir,
str(sample_id) + extra_str + "_minpath_details.txt")
minpath_mps = path.join(minpath_outdir,
str(sample_id) + extra_str + "_minpath.mps")
id_minpath_fh = open(minpath_in, "w")
# Loop over all reactions (which are the index labels in unstrat table
# unless regrouped).
for reaction_id in unstrat_input.index.values:
# Get count of each sequence in sample and write that sequence out
# along with count if non-zero abundance.
reaction_count = unstrat_input.loc[reaction_id, sample_id]
# If 0 then skip.
if reaction_count == 0:
continue
id_minpath_fh.write(reaction_id + "\t" + str(reaction_count) + "\n")
id_minpath_fh.close()
# Run MinPath on this sample.
path2minpath = path.join(path.dirname(path.abspath(__file__)), 'MinPath',
'MinPath12hmp.py')
minpath_cmd = path2minpath + " -any " + minpath_in + " -map " +\
minpath_map + " -report " + minpath_report +\
" -details " + minpath_details + " -mps " + minpath_mps
system_call_check(minpath_cmd, print_out=print_opt)
# Read through MinPath report and keep track of pathways identified
# to be present.
path_present = identify_minpath_present(minpath_report)
# Return list of which pathways are present.
return (path_present)
def castor_hsp_wrapper(tree_path,
trait_tab,
hsp_method,
calc_ci=False,
check_input=False,
ran_seed=None):
'''Wrapper for making system calls to castor_hsp.py Rscript.'''
castor_hsp_script = path.join(get_picrust_project_dir(), 'picrust2',
'Rscripts', 'castor_hsp.R')
# Need to format boolean setting as string for R to read in as argument.
if calc_ci:
calc_ci_setting = "TRUE"
else:
calc_ci_setting = "FALSE"
if check_input:
check_input_setting = "TRUE"
else:
check_input_setting = "FALSE"
# Create temporary directory for writing output files of castor_hsp.R
with TemporaryDirectory() as temp_dir:
output_count_path = path.join(temp_dir, "predicted_counts.txt")
output_ci_path = path.join(temp_dir, "predicted_ci.txt")
hsp_cmd = " ".join([
"Rscript", castor_hsp_script, tree_path, trait_tab, hsp_method,
calc_ci_setting, check_input_setting, output_count_path,
output_ci_path,
str(ran_seed)
])
# Run castor_hsp.R
system_call_check(hsp_cmd)
# Load the output into Table objects
try:
asr_table = pd.read_table(filepath_or_buffer=output_count_path,
sep="\t",
index_col="sequence")
except IOError:
raise ValueError("Cannot read in expected output file" +
output_ci_path)
if calc_ci:
asr_ci_table = pd.read_table(filepath_or_buffer=output_ci_path,
sep="\t",
index_col="sequence")
else:
asr_ci_table = None
# Return list with predicted counts and CIs.
return [asr_table, asr_ci_table]
def test_picrust2_pipeline_script(self):
'''Test that full pipeline can be run successfully with
picrust2_pipeline.py'''
with TemporaryDirectory() as temp_dir:
system_call_check("picrust2_pipeline.py -s " + test_study_seqs +
" -i " + test_seq_abun_tsv + " -o " + temp_dir +
" -r " + test_msa + " -t " + test_tree +
" --custom_trait_tables " + test_known_traits +
" --marker_gene_table " + test_known_marker +
" -o " + temp_dir)
def place_seqs_pipeline(study_fasta, ref_dir, out_tree, threads, out_dir,
chunk_size, print_cmds):
'''Full pipeline for running sequence placement.'''
# Identify reference files to use.
ref_msa, tree, hmm, model = identify_ref_files(ref_dir)
# Run hmmalign to place study sequences into reference MSA.
out_stockholm = path.join(out_dir, "query_align.stockholm")
system_call_check("hmmalign --trim --dna --mapali " + ref_msa +
" --informat FASTA -o " + out_stockholm + " " + hmm +
" " + study_fasta,
print_out=print_cmds)
hmmalign_out = read_stockholm(out_stockholm, clean_char=True)
# Specify split FASTA files to be created.
study_msa_fastafile = path.join(out_dir, "study_seqs_hmmalign.fasta")
ref_msa_fastafile = path.join(out_dir, "ref_seqs_hmmalign.fasta")
ref_seqnames = set(list(read_fasta(ref_msa).keys()))
study_seqnames = set(read_fasta(study_fasta).keys())
ref_hmmalign_subset = {seq: hmmalign_out[seq] for seq in ref_seqnames}
study_hmmalign_subset = {seq: hmmalign_out[seq] for seq in study_seqnames}
write_fasta(ref_hmmalign_subset, ref_msa_fastafile)
write_fasta(study_hmmalign_subset, study_msa_fastafile)
# Run EPA-NG to output .jplace file.
epa_out_dir = path.join(out_dir, "epa_out")
run_epa_ng(tree=tree,
ref_msa_fastafile=ref_msa_fastafile,
study_msa_fastafile=study_msa_fastafile,
model=model,
chunk_size=chunk_size,
threads=threads,
out_dir=epa_out_dir,
print_cmds=print_cmds)
jplace_outfile = path.join(epa_out_dir, "epa_result_parsed.jplace")
gappa_jplace_to_newick(jplace_file=jplace_outfile,
outfile=out_tree,
print_cmds=print_cmds)
def gappa_jplace_to_newick(jplace_file: str, outfile: str, print_cmds=False):
'''System call to gappa binary to convert jplace object to newick
treefile (with specified filename).'''
gappa_out_dir = path.dirname(jplace_file)
# Run gappa to convert jplace to newick.
system_call_check("gappa analyze graft --jplace-path " + jplace_file +
" --fully-resolve --out-dir " + gappa_out_dir,
print_out=print_cmds)
# Expected name of output newick file.
newick_file = jplace_file.replace(".jplace", ".newick")
# Rename newick file to be specified outfile.
system_call_check("mv " + newick_file + " " + outfile,
print_out=print_cmds)
def run_epa_ng(tree: str,
ref_msa_fastafile: str,
study_msa_fastafile: str,
out_dir: str,
chunk_size=5000,
threads=1,
print_cmds=False):
'''Run EPA-NG on specified tree, reference MSA, and study sequence MSA.
Will opath.joinutput a .jplace file in out_dir.'''
make_output_dir(out_dir)
system_call_check("epa-ng --tree " + tree + " --ref-msa " +
ref_msa_fastafile + " --query " + study_msa_fastafile +
" --chunk-size " + str(chunk_size) + " -T " +
str(threads) + " -w " + out_dir,
print_out=print_cmds)
def test_full_pipeline_biom(self):
'''Test that full pipeline can be run without error with
BIOM sequence abundance table.'''
with TemporaryDirectory() as temp_dir:
out_tree = path.join(temp_dir, "out.tre")
system_call_check("place_seqs.py -s " + test_study_seqs + " -r " +\
test_msa + " -t " + test_tree + " -o " +\
out_tree)
hsp_out_prefix = path.join(temp_dir, "hsp_out")
hsp_out_prefix_marker = path.join(temp_dir, "hsp_out_marker")
system_call_check("hsp.py -t " + out_tree +\
" --observed_trait_table " + test_known_traits + " -n -c " +\
"-o " + hsp_out_prefix)
system_call_check("hsp.py -t " + out_tree +\
" --observed_trait_table " + test_known_marker + " -n -c " +\
"-o " + hsp_out_prefix_marker)
traits_predict = path.join(temp_dir, hsp_out_prefix +\
".tsv")
marker_predict = path.join(temp_dir, hsp_out_prefix_marker +\
".tsv")
metagenome_out = path.join(temp_dir, "meta_out")
system_call_check("metagenome_pipeline.py -i " + test_seq_abun_biom +\
" -f " + traits_predict + " -m " +
marker_predict + " -o " + metagenome_out)
metagenome_outfile = path.join(metagenome_out,
"pred_metagenome_strat.tsv")
minpath_out = path.join(temp_dir, "minpath_out")
system_call_check("run_minpath.py -i " + metagenome_outfile +\
" -m " + minpath_map + " -o " + minpath_out)
def castor_hsp_loocv_wrapper(tree_path,
trait_table_path,
tips_path,
hsp_method,
expected_out_path,
predicted_out_path,
metrics_out_path,
num_cores=1):
'''Runs the castor_hsp_loocv.R Rscript and writes out result tables'''
castor_loocv_hsp_script_fp = path.join(get_picrust_project_dir(),
'picrust2', 'Rscripts',
'castor_hsp_loocv.R')
loocv_cmd = " ".join([
"Rscript", castor_loocv_hsp_script_fp, tree_path, trait_table_path,
tips_path, hsp_method, expected_out_path, predicted_out_path,
metrics_out_path,
str(num_cores)
])
# Run castor_hsp_loocv.R here
system_call_check(loocv_cmd)
def full_pipeline(study_fasta, input_table, output_folder, processes, ref_dir,
in_traits, custom_trait_tables, marker_gene_table,
pathway_map, rxn_func, no_pathways, regroup_map, no_regroup,
stratified, max_nsti, min_reads, min_samples, hsp_method,
min_align, skip_nsti, skip_minpath, no_gap_fill, coverage,
per_sequence_contrib, wide_table, skip_norm,
remove_intermediate, verbose):
'''Function that contains wrapper commands for full PICRUSt2 pipeline.
Descriptions of all of these input arguments/options are given in the
picrust2_pipeline.py script.'''
# Throw warning if --per_sequence_contrib set but --stratified unset.
if per_sequence_contrib and not stratified:
print(
"\nThe option --per_sequence_contrib was set, but not the option "
"--stratified. This means that a stratified pathway table will "
"be output only (i.e. a stratified metagenome table will NOT "
"be output).\n",
file=sys.stderr)
out_tree = path.join(output_folder, "out.tre")
if custom_trait_tables is None:
# Check that specified functional categories are allowed.
FUNC_TRAIT_OPTIONS = ['COG', 'EC', 'KO', 'PFAM', 'TIGRFAM']
funcs = in_traits.split(",")
for func in funcs:
if func not in FUNC_TRAIT_OPTIONS:
sys.exit("Error - specified category " + func + " is not " +
"one of the default categories.")
func_tables = default_tables
else:
# Split paths to input custom trait tables and take the basename to be
# the function id.
funcs = []
func_tables = {}
for custom in custom_trait_tables.split(","):
func_id = path.splitext(path.basename(custom))[0]
funcs.append(func_id)
func_tables[func_id] = custom
# Add reaction function to be in set of gene families if it is not already
# and as long as pathways are also to be predicted.
if rxn_func not in funcs and not no_pathways:
orig_rxn_func = rxn_func
rxn_func = path.splitext(path.basename(rxn_func))[0]
funcs.append(rxn_func)
if rxn_func not in func_tables:
func_tables[rxn_func] = orig_rxn_func
if not skip_norm:
# Append marker as well, since this also needs to be run.
funcs.append("marker")
func_tables["marker"] = marker_gene_table
# Check that all input files exist.
ref_msa, tree, hmm, model = identify_ref_files(ref_dir)
files2check = [study_fasta, input_table, ref_msa, tree, hmm, model] + list(
func_tables.values())
if not no_pathways:
files2check.append(pathway_map)
# Throw warning if default pathway mapfile used with non-default
# reference files.
if pathway_map == default_pathway_map and ref_dir != default_ref_dir:
print(
"Warning - non-default reference files specified with "
"default pathway mapfile of prokaryote-specific MetaCyc "
"pathways (--pathway_map option). This usage may be "
"unintended.",
file=sys.stderr)
if not no_regroup:
files2check.append(regroup_map)
# This will throw an error if any input files are not found.
check_files_exist(files2check)
# Check that sequence names in FASTA overlap with input table.
check_overlapping_seqs(study_fasta, input_table, verbose)
if path.exists(output_folder):
sys.exit("Stopping since output directory " + output_folder +
" already exists.")
# Make output folder.
make_output_dir(output_folder)
if verbose:
print("Placing sequences onto reference tree", file=sys.stderr)
# Define folders for intermediate files (unless --remove_intermediate set).
if remove_intermediate:
place_seqs_intermediate = ""
pathways_intermediate = ""
else:
intermediate_dir = path.join(output_folder, "intermediate")
make_output_dir(intermediate_dir)
place_seqs_intermediate = path.join(intermediate_dir, "place_seqs")
pathways_intermediate = path.join(intermediate_dir, "pathways")
# Run place_seqs.py.
place_seqs_cmd = [
"place_seqs.py", "--study_fasta", study_fasta, "--ref_dir", ref_dir,
"--out_tree", out_tree, "--processes",
str(processes), "--intermediate", place_seqs_intermediate,
"--min_align",
str(min_align), "--chunk_size",
str(5000)
]
if verbose:
place_seqs_cmd.append("--verbose")
system_call_check(place_seqs_cmd,
print_command=verbose,
print_stdout=verbose,
print_stderr=True)
if verbose:
print("Finished placing sequences on output tree: " + out_tree,
file=sys.stderr)
# Get predictions for all specified functions and keep track of outfiles.
predicted_funcs = {}
if not skip_norm:
# Make sure marker database is first in the list. This is because this will
# be run on a single core and so will be easier to identify any errors
# if the program exits when working on this function type.
funcs.insert(0, funcs.pop(funcs.index("marker")))
for func in funcs:
# Change output filename for NSTI and non-NSTI containing files.
hsp_outfile = path.join(output_folder, func + "_predicted")
if (func == "marker" and not skip_nsti) or (skip_norm
and not skip_nsti):
hsp_outfile = hsp_outfile + "_and_nsti.tsv.gz"
else:
hsp_outfile = hsp_outfile + ".tsv.gz"
# Keep track of output filename for next step of pipeline.
predicted_funcs[func] = hsp_outfile
# Run hsp.py for each function database.
hsp_cmd = [
"hsp.py", "--tree", out_tree, "--output", hsp_outfile,
"--observed_trait_table", func_tables[func], "--hsp_method",
hsp_method, "--seed", "100"
]
# Add flags to command if specified.
if (func == "marker" and not skip_nsti) or (skip_norm
and not skip_nsti):
hsp_cmd.append("--calculate_NSTI")
# Run marker on only 1 processor.
if func == "marker":
hsp_cmd += ["--processes", "1"]
else:
hsp_cmd += ["--processes", str(processes)]
if verbose:
hsp_cmd.append("--verbose")
system_call_check(hsp_cmd,
print_command=verbose,
print_stdout=verbose,
print_stderr=True)
# Now run metagenome pipeline commands.
# Inititalize dictionary of function names --> metagenome output files.
func_output = {}
# Loop over each function again and run metagenome pipeline.
for func in funcs:
if func == "marker":
continue
if verbose:
print("Running metagenome pipeline for " + func, file=sys.stderr)
func_output_dir = path.join(output_folder, func + "_metagenome_out")
metagenome_pipeline_cmd = [
"metagenome_pipeline.py", "--input", input_table, "--function",
predicted_funcs[func], "--min_reads",
str(min_reads), "--min_samples",
str(min_samples), "--out_dir", func_output_dir
]
# Initialize two-element list as value for each function.
# First value will be unstratified output and second will be
# stratified output.
func_output[func] = [None, None]
func_output[func][0] = path.join(func_output_dir,
"pred_metagenome_unstrat.tsv.gz")
if wide_table:
metagenome_pipeline_cmd.append("--wide_table")
if not skip_nsti:
metagenome_pipeline_cmd += ["--max_nsti", str(max_nsti)]
if skip_norm:
metagenome_pipeline_cmd.append("--skip_norm")
else:
metagenome_pipeline_cmd += ["--marker", predicted_funcs["marker"]]
if stratified:
metagenome_pipeline_cmd.append("--strat_out")
if wide_table:
func_output[func][1] = path.join(
func_output_dir, "pred_metagenome_strat.tsv.gz")
else:
func_output[func][1] = path.join(
func_output_dir, "pred_metagenome_contrib.tsv.gz")
system_call_check(metagenome_pipeline_cmd,
print_command=verbose,
print_stdout=verbose,
print_stderr=True)
# Now infer pathway abundances and coverages unless --no_pathways set.
pathway_outfiles = None
if not no_pathways:
path_output_dir = path.join(output_folder, "pathways_out")
if verbose:
print("Inferring pathways from predicted " + rxn_func)
# Determine whether stratified or unstratified table should be input.
if not stratified or per_sequence_contrib:
rxn_input_metagenome = func_output[rxn_func][0]
else:
rxn_input_metagenome = func_output[rxn_func][1]
pathway_pipeline_cmd = [
"pathway_pipeline.py", "--input", rxn_input_metagenome,
"--out_dir", path_output_dir, "--map", pathway_map,
"--intermediate", pathways_intermediate, "--proc",
str(processes)
]
if no_gap_fill:
pathway_pipeline_cmd.append("--no_gap_fill")
if skip_minpath:
pathway_pipeline_cmd.append("--skip_minpath")
if coverage:
pathway_pipeline_cmd.append("--coverage")
if no_regroup:
pathway_pipeline_cmd.append("--no_regroup")
else:
pathway_pipeline_cmd += ["--regroup_map", regroup_map]
if wide_table:
pathway_pipeline_cmd.append("--wide_table")
if per_sequence_contrib:
pathway_pipeline_cmd.append("--per_sequence_contrib")
if skip_norm:
norm_sequence_abun = input_table
else:
norm_sequence_abun = path.join(output_folder,
rxn_func + "_metagenome_out",
"seqtab_norm.tsv.gz")
pathway_pipeline_cmd += ["--per_sequence_abun", norm_sequence_abun]
pathway_pipeline_cmd += [
"--per_sequence_function", predicted_funcs[rxn_func]
]
if verbose:
pathway_pipeline_cmd.append("--verbose")
system_call_check(pathway_pipeline_cmd,
print_command=verbose,
print_stdout=False,
print_stderr=True)
if verbose:
print("Wrote predicted pathway abundances and coverages to " +
path_output_dir,
file=sys.stderr)
# Keep track of output filenames if this function is being used in
# a non-default way (e.g. with a QIIME2 plugin).
pathway_outfiles = {}
pathway_outfiles["unstrat_abun"] = path.join(
path_output_dir, "path_abun_unstrat.tsv.gz")
pathway_outfiles["unstrat_cov"] = path.join(path_output_dir,
"path_cov_unstrat.tsv.gz")
pathway_outfiles["strat_abun"] = None
pathway_outfiles["strat_cov"] = None
if stratified or per_sequence_contrib:
if wide_table:
pathway_outfiles["strat_abun"] = path.join(
path_output_dir, "path_abun_strat.tsv.gz")
if per_sequence_contrib:
pathway_outfiles["strat_cov"] = path.join(
path_output_dir, "path_cov_strat.tsv.gz")
else:
pathway_outfiles["strat_abun"] = path.join(
path_output_dir, "path_abun_contrib.tsv.gz")
if per_sequence_contrib:
pathway_outfiles["strat_cov"] = path.join(
path_output_dir, "path_cov_contrib.tsv.gz")
return (func_output, pathway_outfiles)
def place_seqs_pipeline(study_fasta,
ref_dir,
out_tree,
threads,
out_dir,
min_align,
chunk_size,
verbose):
'''Full pipeline for running sequence placement.'''
# Throw error if there is a space in the study FASTA filepath.
if " " in study_fasta:
sys.exit("Stopping - remove the space from the input FASTA filepath.")
# Identify reference files to use.
ref_msa, tree, hmm, model = identify_ref_files(ref_dir)
# Run hmmalign to place study sequences into reference MSA.
out_stockholm = path.join(out_dir, "query_align.stockholm")
system_call_check("hmmalign --trim --dna --mapali " +
ref_msa + " --informat FASTA -o " +
out_stockholm + " " + hmm + " " + study_fasta,
print_command=verbose, print_stdout=verbose,
print_stderr=verbose)
hmmalign_out = read_stockholm(out_stockholm, clean_char=True)
# Specify split FASTA files to be created.
study_msa_fastafile = path.join(out_dir, "study_seqs_hmmalign.fasta")
ref_msa_fastafile = path.join(out_dir, "ref_seqs_hmmalign.fasta")
ref_seqnames = set(list(read_fasta(ref_msa).keys()))
study_seqs = read_fasta(study_fasta)
ref_hmmalign_subset = {seq: hmmalign_out[seq] for seq in ref_seqnames}
study_hmmalign_subset = check_alignments(raw_seqs=study_seqs,
aligned_seqs=hmmalign_out,
min_align=min_align,
verbose=verbose)
write_fasta(ref_hmmalign_subset, ref_msa_fastafile)
write_fasta(study_hmmalign_subset, study_msa_fastafile)
# Run EPA-ng to place input sequences and output JPLACE file.
epa_out_dir = path.join(out_dir, "epa_out")
run_epa_ng(tree=tree,
ref_msa_fastafile=ref_msa_fastafile,
study_msa_fastafile=study_msa_fastafile,
model=model,
chunk_size=chunk_size,
threads=threads,
out_dir=epa_out_dir,
print_cmds=verbose)
jplace_outfile = path.join(epa_out_dir, "epa_result_parsed.jplace")
gappa_jplace_to_newick(jplace_file=jplace_outfile,
outfile=out_tree,
print_cmds=verbose)
def place_seqs_pipeline(study_fasta, ref_msa, tree, hmm, out_tree,
alignment_tool, threads, out_dir, chunk_size,
print_cmds):
'''Full pipeline for running sequence placement.'''
if alignment_tool == "hmmalign":
out_stockholm = path.join(out_dir, "query_align.stockholm")
system_call_check("hmmalign --trim --dna --mapali " + ref_msa +
" --informat FASTA -o " + out_stockholm + " " + hmm +
" " + study_fasta,
print_out=print_cmds)
hmmalign_out = read_stockholm(out_stockholm, clean_char=True)
# Specify split FASTA files to be created.
study_msa_fastafile = path.join(out_dir, "study_seqs_hmmalign.fasta")
ref_msa_fastafile = path.join(out_dir, "ref_seqs_hmmalign.fasta")
ref_seqnames = set(list(read_fasta(ref_msa).keys()))
study_seqnames = set(read_fasta(study_fasta).keys())
ref_hmmalign_subset = {seq: hmmalign_out[seq] for seq in ref_seqnames}
study_hmmalign_subset = {
seq: hmmalign_out[seq]
for seq in study_seqnames
}
write_fasta(ref_hmmalign_subset, ref_msa_fastafile)
write_fasta(study_hmmalign_subset, study_msa_fastafile)
elif alignment_tool == "papara":
# Read in ref seqs FASTA as a dict.
ref_msa = read_fasta(ref_msa)
# Run PaPaRa to place study sequences and read in Phylip file.
papara_out = run_papara(tree=tree,
ref_msa=ref_msa,
study_fasta=study_fasta,
out_dir=out_dir,
threads=threads,
print_cmds=print_cmds)
# Specify split FASTA files to be created.
study_msa_fastafile = path.join(out_dir, "study_seqs_papara.fasta")
ref_msa_fastafile = path.join(out_dir, "ref_seqs_papara.fasta")
# Split PaPaRa output into two FASTA files containing study and reference
# sequences respectively.
split_ref_study_papara(papara_out=papara_out,
ref_seqnames=set(list(ref_msa.keys())),
study_fasta=study_msa_fastafile,
ref_fasta=ref_msa_fastafile)
# Run EPA-NG to output .jplace file.
epa_out_dir = path.join(out_dir, "epa_out")
run_epa_ng(tree=tree,
ref_msa_fastafile=ref_msa_fastafile,
study_msa_fastafile=study_msa_fastafile,
chunk_size=chunk_size,
threads=threads,
out_dir=epa_out_dir,
print_cmds=print_cmds)
jplace_outfile = path.join(epa_out_dir, "epa_result.jplace")
gappa_jplace_to_newick(jplace_file=jplace_outfile,
outfile=out_tree,
print_cmds=print_cmds)
def custom_tree_pipeline(table: biom.Table,
tree: skbio.TreeNode,
threads: int = 1,
hsp_method: str = "mp",
max_nsti: float = 2.0,
edge_exponent: float = 0.5,
skip_minpath: bool = False,
no_gap_fill: bool = False,
skip_norm: bool = False,
highly_verbose: bool = False) -> (biom.Table,
biom.Table,
biom.Table):
# Run pipeline in temporary directory so that files are not saved locally.
with TemporaryDirectory() as temp_dir:
# Need to write out BIOM table and newick tree to be used in pipeline.
# Write out biom table:
biom_infile = path.join(temp_dir, "intable.biom")
with biom.util.biom_open(biom_infile, 'w') as out_biom:
table.to_hdf5(h5grp=out_biom,
generated_by="PICRUSt2 QIIME 2 Plugin")
# Write out newick tree.
newick_infile = path.join(temp_dir, "placed_seqs.tre")
tree.write(newick_infile, format="newick")
picrust2_out = path.join(temp_dir, "picrust2_out")
print("Running the below commands:", file=sys.stderr)
# Run hidden-state prediction step (on 16S, EC, and KO tables
# separately.
hsp_out_16S = path.join(picrust2_out, "16S_predicted.tsv.gz")
hsp_out_16S_cmd = "hsp.py -i 16S " + \
" -t " + newick_infile + \
" -p 1 " + \
" -n " + \
" -o " + hsp_out_16S + \
" -m " + hsp_method + \
" -e " + str(edge_exponent)
hsp_out_EC = path.join(picrust2_out, "EC_predicted.tsv.gz")
hsp_out_EC_cmd = "hsp.py -i EC " + \
" -t " + newick_infile + \
" -p " + str(threads) + \
" -n " + \
" -o " + hsp_out_EC + \
" -m " + hsp_method + \
" -e " + str(edge_exponent)
hsp_out_KO = path.join(picrust2_out, "KO_predicted.tsv.gz")
hsp_out_KO_cmd = "hsp.py -i KO " + \
" -t " + newick_infile + \
" -p " + str(threads) + \
" -n " + \
" -o " + hsp_out_KO + \
" -m " + hsp_method + \
" -e " + str(edge_exponent)
if highly_verbose:
hsp_out_16S_cmd += " --verbose"
hsp_out_EC_cmd += " --verbose"
hsp_out_KO_cmd += " --verbose"
if not skip_norm:
system_call_check(hsp_out_16S_cmd,
print_command=True,
print_stdout=highly_verbose,
print_stderr=True)
system_call_check(hsp_out_EC_cmd,
print_command=True,
print_stdout=highly_verbose,
print_stderr=True)
system_call_check(hsp_out_KO_cmd,
print_command=True,
print_stdout=highly_verbose,
print_stderr=True)
# Run metagenome pipeline step.
EC_metagenome_out = path.join(picrust2_out, "EC_metagenome_out")
KO_metagenome_out = path.join(picrust2_out, "KO_metagenome_out")
EC_metagenome_cmd = "metagenome_pipeline.py -i " + biom_infile + \
" -f " + hsp_out_EC + \
" -o " + EC_metagenome_out + \
" --max_nsti " + str(max_nsti)
KO_metagenome_cmd = "metagenome_pipeline.py -i " + biom_infile + \
" -f " + hsp_out_KO + \
" -o " + KO_metagenome_out + \
" --max_nsti " + str(max_nsti)
if skip_norm:
EC_metagenome_cmd += " --skip_norm"
KO_metagenome_cmd += " --skip_norm"
else:
EC_metagenome_cmd += " -m " + hsp_out_16S
KO_metagenome_cmd += " -m " + hsp_out_16S
system_call_check(EC_metagenome_cmd, print_command=True,
print_stdout=highly_verbose,
print_stderr=True)
system_call_check(KO_metagenome_cmd, print_command=True,
print_stdout=highly_verbose,
print_stderr=True)
EC_out = path.join(EC_metagenome_out, "pred_metagenome_unstrat.tsv.gz")
KO_out = path.join(KO_metagenome_out, "pred_metagenome_unstrat.tsv.gz")
# Run pathway inference step.
pathways_out = path.join(picrust2_out, "pathways_out")
pathabun_out = path.join(pathways_out, "path_abun_unstrat.tsv.gz")
pathway_pipeline_cmd = "pathway_pipeline.py -i " + EC_out + \
" -o " + pathways_out + \
" -p " + str(threads)
if skip_minpath:
pathway_pipeline_cmd += " --skip_minpath"
if no_gap_fill:
pathway_pipeline_cmd += " --no_gap_fill"
if highly_verbose:
pathway_pipeline_cmd += " --verbose"
system_call_check(pathway_pipeline_cmd, print_command=True,
print_stdout=highly_verbose,
print_stderr=True)
# Read in output unstratified metagenome tables and return as BIOM
# objects.
ko_biom = biom.load_table(KO_out)
ec_biom = biom.load_table(EC_out)
pathabun_biom = biom.load_table(pathabun_out)
return ko_biom, ec_biom, pathabun_biom
def minpath_wrapper(sample_id, strat_input, minpath_map, out_dir,
print_opt=False):
'''Read in sample_id, gene family table, and out_dir, and run MinPath based
on the gene family abundances. Returns both unstratified and stratified
pathway abundances as dictionaries in a list.'''
# Get gene family abundances summed over all sequences for this sample.
unstrat_input = strat_to_unstrat_counts(strat_input)
# Define MinPath input and outout filenames.
minpath_in = path.join(out_dir, sample_id + "_minpath_in.txt")
minpath_report = path.join(out_dir, sample_id + "_minpath_report.txt")
minpath_details = path.join(out_dir, sample_id + "_minpath_details.txt")
minpath_mps = path.join(out_dir, sample_id + "_minpath.mps")
minpath_output = open(path.join(out_dir, sample_id + "_minpath_out.txt"),
"w")
id_minpath_fh = open(minpath_in, "w")
# Loop over all functions (which are the index labels in unstrat table).
for func_id in unstrat_input.index.values:
# Get count of each sequence in sample and write that sequence out
# along with count if non-zero abundance.
func_count = unstrat_input.loc[func_id, sample_id]
# If 0 then skip.
if func_count == 0:
continue
id_minpath_fh.write(func_id + "\t" + str(func_count) + "\n")
id_minpath_fh.close()
# Run MinPath on this sample.
path2minpath = path.join(get_picrust_project_dir(), 'MinPath',
'MinPath12hmp.py')
minpath_cmd = path2minpath + " -any " + minpath_in + " -map " +\
minpath_map + " -report " + minpath_report +\
" -details " + minpath_details + " -mps " + minpath_mps
system_call_check(minpath_cmd, print_out=print_opt,
stdout=minpath_output)
# Read through MinPath report and keep track of pathways identified
# to be present.
path_present = identify_minpath_present(minpath_report)
# Now read in details file and take abundance of pathway to be
# mean of top 1/2 most abundant gene families.
# Abundances of 0 will be added in for gene families not found.
gf_abundances, gf_ids = parse_minpath_details(minpath_details, path_present)
# Initialize series and dataframe that will contain pathway abundances.
unstrat_abun = pd.Series()
strat_abun = pd.DataFrame(columns=["pathway", "sequence", sample_id])
strat_abun = strat_abun.set_index(["pathway", "sequence"])
# Loop through all pathways present and get mean of 1/2 most abundant.
for pathway in gf_abundances.keys():
# Like HUMAnN2, sort enzyme reactions, take second half, and get
# their mean abundance.
# First get indices of sorted list.
sorted_index = list(np.argsort(gf_abundances[pathway]))
sorted_gf_abundances = [gf_abundances[pathway][i] for i in sorted_index]
sorted_gf_ids = [gf_ids[pathway][i] for i in sorted_index]
# Take second half of gene family abundances and ids lists.
half_i = int(len(sorted_gf_abundances) / 2)
gf_abundances_subset = sorted_gf_abundances[half_i:]
gf_ids_subset = sorted_gf_ids[half_i:]
# Take mean for unstratified pathway abundance.
unstrat_abun[pathway] = sum(gf_abundances_subset)/len(gf_abundances_subset)
# Get stratified pathway abundances by sequences.
strat_path_abun = path_abun_by_seq(strat_input,
gf_ids_subset,
sum(gf_abundances_subset),
unstrat_abun[pathway])
# Remove rows that are all 0.
strat_path_abun[strat_path_abun[sample_id] > 0]
# Add pathway as new column.
strat_path_abun["pathway"] = [pathway]*strat_path_abun.shape[0]
strat_path_abun.set_index("pathway", append=True, inplace=True)
# Changes levels of index labels.
strat_path_abun = strat_path_abun.reorder_levels(["pathway",
"sequence"])
strat_abun = pd.concat([strat_abun, strat_path_abun], levels=["pathway", "sequence"])
# Return unstratified and stratified abundances.
# Note that the stratified abundances are converted to a series.
return([unstrat_abun, strat_abun[sample_id]])
def castor_hsp_wrapper(tree_path,
trait_tab,
hsp_method,
calc_ci=False,
check_input=False,
ran_seed=None,
verbose=False):
'''Wrapper for making system calls to castor_hsp.py Rscript.'''
castor_hsp_script = path.join(path.dirname(path.abspath(__file__)),
'Rscripts', 'castor_hsp.R')
# Need to format boolean setting as string for R to read in as argument.
if calc_ci:
calc_ci_setting = "TRUE"
else:
calc_ci_setting = "FALSE"
if check_input:
check_input_setting = "TRUE"
else:
check_input_setting = "FALSE"
# Create temporary directory for writing output files of castor_hsp.R
with TemporaryDirectory() as temp_dir:
output_count_path = path.join(temp_dir, "predicted_counts.txt")
output_ci_path = path.join(temp_dir, "predicted_ci.txt")
hsp_cmd = " ".join([
"Rscript", castor_hsp_script, tree_path, trait_tab, hsp_method,
calc_ci_setting, check_input_setting, output_count_path,
output_ci_path,
str(ran_seed)
])
# Run castor_hsp.R
system_call_check(hsp_cmd,
print_command=verbose,
print_stdout=verbose,
print_stderr=verbose)
# Load the output into Table objects
try:
asr_table = pd.read_csv(filepath_or_buffer=output_count_path,
sep="\t",
dtype={'sequence': str})
asr_table.set_index('sequence', drop=True, inplace=True)
except IOError:
raise ValueError("Cannot read in expected output file" +
output_ci_path)
if calc_ci:
asr_ci_table = pd.read_csv(filepath_or_buffer=output_ci_path,
sep="\t",
dtype={'sequence': str})
asr_ci_table.set_index('sequence', drop=True, inplace=True)
else:
asr_ci_table = None
# Return list with predicted counts and CIs.
return [asr_table, asr_ci_table]
def custom_tree_pipeline(
table: biom.Table,
tree: skbio.TreeNode,
threads: int = 1,
hsp_method: str = "mp",
max_nsti: float = 2.0) -> (biom.Table, biom.Table, biom.Table):
# Run pipeline in temporary directory so that files are not saved locally.
with TemporaryDirectory() as temp_dir:
# Need to write out BIOM table and newick tree to be used in pipeline.
# Write out biom table:
biom_infile = path.join(temp_dir, "intable.biom")
with biom.util.biom_open(biom_infile, 'w') as out_biom:
table.to_hdf5(h5grp=out_biom,
generated_by="PICRUSt2 QIIME2 Plugin")
# Write out newick tree.
newick_infile = path.join(temp_dir, "placed_seqs.tre")
tree.write(newick_infile, format="newick")
picrust2_out = path.join(temp_dir, "picrust2_out")
print("Running the below commands:", file=sys.stderr)
# Run hidden-state prediction step (on 16S, EC, and KO tables
# separately.
hsp_out_16S = path.join(picrust2_out, "16S_predicted.tsv.gz")
system_call_check("hsp.py -i 16S " + " -t " + newick_infile +
" -p 1 " + " -n " + "-o " + hsp_out_16S + " -m " +
hsp_method,
print_out=True)
hsp_out_EC = path.join(picrust2_out, "EC_predicted.tsv.gz")
system_call_check("hsp.py -i EC " + " -t " + newick_infile + " -p " +
str(threads) + " -o " + hsp_out_EC + " -m " +
hsp_method,
print_out=True)
hsp_out_KO = path.join(picrust2_out, "KO_predicted.tsv.gz")
system_call_check("hsp.py -i KO " + " -t " + newick_infile + " -p " +
str(threads) + " -o " + hsp_out_KO + " -m " +
hsp_method,
print_out=True)
# Run metagenome pipeline step.
EC_metagenome_out = path.join(picrust2_out, "EC_metagenome_out")
system_call_check("metagenome_pipeline.py -i " + biom_infile + " -m " +
hsp_out_16S + " -f " + hsp_out_EC + " -o " +
EC_metagenome_out + " --max_nsti " + str(max_nsti),
print_out=True)
KO_metagenome_out = path.join(picrust2_out, "KO_metagenome_out")
system_call_check("metagenome_pipeline.py -i " + biom_infile + " -m " +
hsp_out_16S + " -f " + hsp_out_KO + " -o " +
KO_metagenome_out + " --max_nsti " + str(max_nsti),
print_out=True)
EC_out = path.join(EC_metagenome_out, "pred_metagenome_unstrat.tsv.gz")
KO_out = path.join(KO_metagenome_out, "pred_metagenome_unstrat.tsv.gz")
# Run pathway inference step.
pathways_out = path.join(picrust2_out, "pathways_out")
pathabun_out = path.join(pathways_out, "path_abun_unstrat.tsv.gz")
system_call_check("pathway_pipeline.py -i " + EC_out + " -o " +
pathways_out + " -p " + str(threads),
print_out=True)
# Read in output unstratified metagenome tables and return as BIOM
# objects.
ko_biom = biom.load_table(KO_out)
ec_biom = biom.load_table(EC_out)
pathabun_biom = biom.load_table(pathabun_out)
return ko_biom, ec_biom, pathabun_biom
def minpath_wrapper(sample_id,
unstrat_input,
minpath_map,
out_dir,
print_opt=False,
extra_str=""):
'''Run MinPath based on gene abundances in a single sample. Will return
the abundances of gene families within each identified pathway.'''
# Make output directory for MinPath intermediate files.
make_output_dir(path.join(out_dir, "minpath_running"))
# Define MinPath input and output filenames.
minpath_in = path.join(out_dir, "minpath_running",
sample_id + extra_str + "_minpath_in.txt")
minpath_report = path.join(out_dir, "minpath_running",
sample_id + extra_str + "_minpath_report.txt")
minpath_details = path.join(out_dir, "minpath_running",
sample_id + extra_str + "_minpath_details.txt")
minpath_mps = path.join(out_dir, "minpath_running",
sample_id + extra_str + "_minpath.mps")
minpath_output = open(path.join(out_dir, sample_id + "_minpath_out.txt"),
"w")
id_minpath_fh = open(minpath_in, "w")
# Inititalize dictionary for keeping track of reaction abundances.
reaction_abun = defaultdict(int)
# Loop over all reactions (which are the index labels in unstrat table
# unless regrouped).
for reaction_id in unstrat_input.index.values:
# Get count of each sequence in sample and write that sequence out
# along with count if non-zero abundance.
reaction_count = unstrat_input.loc[reaction_id, sample_id]
# If 0 then skip.
if reaction_count == 0:
continue
id_minpath_fh.write(reaction_id + "\t" + str(reaction_count) + "\n")
reaction_abun[reaction_id] = reaction_count
id_minpath_fh.close()
# Run MinPath on this sample.
path2minpath = path.join(get_picrust_project_dir(), 'picrust2', 'MinPath',
'MinPath12hmp.py')
minpath_cmd = path2minpath + " -any " + minpath_in + " -map " +\
minpath_map + " -report " + minpath_report +\
" -details " + minpath_details + " -mps " + minpath_mps
system_call_check(minpath_cmd, print_out=print_opt, stdout=minpath_output)
# Read through MinPath report and keep track of pathways identified
# to be present.
path_present = identify_minpath_present(minpath_report)
# Return list of which pathways are present and the abundances of all gene
# families.
return (path_present, reaction_abun)
