def cazy_parser(tsv, namemap, output, summary_method='best', min_identity=60, min_bitscore=0, min_length=60, max_evalue=0.000001, top_fraction=1, max_hits=10, table_name="cazy"):
"""Parse BLAST hits from CAZy reference database.
The BLAST hits are assumed to be sorted by query with decreasing bitscores (best alignment first):
\b
sort -k1,1 -k12,12rn tsv > sorted_tsv
Expected columns in the CAZy database:
\b
cazy_gene
cazy_family
cazy_class
cazy_ec
For a given gene match, all possible ECs are returned in a single line separated by '|'.
Args:
tsv (str): blast hits file path in default tabular format
namemap (str): sqlite database file path
output (str): :py:class:`click.File` in write mode
summary_method (str): either 'majority' or 'best'; summary method for annotating ORFs; when majority and there is no majority, best is used
min_identity (int): minimum allowable percent ID of BLAST hit
min_bitscore (int): minimum allowable bitscore of BLAST hit; 0 disables
min_length (int): minimum allowable BLAST alignment length
max_evalue (float): maximum allowable e-value of BLAST hit
top_fraction (float): filters ORF BLAST hits before finding majority by only keep hits within this fraction, e.g. 0.98, of the highest bitscore
max_hits (int): maximum number of BLAST hits to consider when summarizing ORFs as a majority
table_name (str): table name within namemap database; expected columns are listed above
"""
logging.info("Parsing %s" % tsv)
if top_fraction == 1:
top_fraction = None
print("contig", "orf", "cazy_gene", "cazy_family", "cazy_class", "cazy_ec",
"%s_evalue" % table_name, "%s_bitscore" % table_name, sep="\t", file=output)
with contextlib.closing(sqlite3.connect(namemap)) as conn, gzopen(tsv) as blast_tab_fh:
cursor = conn.cursor()
for query, qgroup in groupby(blast_tab_fh, key=lambda x: x.partition("\t")[0]):
contig_name, _, orf_idx = query.rpartition("_")
hit_id, evalue, bitscore = get_hit_from_blast_group(qgroup, max_hits, top_fraction, min_length, min_identity, max_evalue, min_bitscore, summary_method)
cazy_gene = "NA"
cazy_family = "NA"
cazy_class = "NA"
cazy_ec = "NA"
# everything could have been filtered out due to user constraints
if hit_id:
cursor.execute('SELECT cazy_gene, cazy_family, cazy_class, cazy_ec \
FROM %s \
WHERE cazy_gene="%s"' % (table_name, hit_id))
cazy_gene, cazy_family, cazy_class, cazy_ec = cursor.fetchone()
print(contig_name, "%s_%s" % (contig_name, orf_idx), cazy_gene, cazy_family,
cazy_class, cazy_ec, evalue, bitscore, sep="\t", file=output)
logging.info("Complete")
def parse_blast_results_with_tree(blast_tab,
name_map,
summary_method,
tree,
min_identity=70,
min_bitscore=0,
min_length=60,
max_evalue=0.000001,
max_hits_per_orf=10,
top_fraction_of_hits=None,
table_name="refseq",
lca_threshold=1):
"""Parse BLAST results (-outfmt 6), filter, and aggregate ORF taxonomies.
Args:
blast_tab (str): file path to blast TSV file
name_map (dict): dict of tuples from parse_tree_annotation
summary_method (dict): method of ORF annotation selection
lca_threshold (float): the first parent above this fraction of representation (its count is
greater than the total * lca_threshold)
Returns:
dict: dict of dicts where first key is contig name, inner key is ORF ID; values are tuple of
protein function, taxonomy ID, bitscore, evalue
Raises:
AssertionError when ORF summary method is not supported (['lca', 'best', 'majority'])
"""
# allowing 1 and 0 to disable
if top_fraction_of_hits == 1:
top_fraction_of_hits = None
assert summary_method in ["lca", "best", "majority"]
contigs = defaultdict(dict)
with contextlib.closing(sqlite3.connect(name_map)) as conn, gzopen(
blast_tab) as blast_tab_fh:
cursor = conn.cursor()
# group hits by ORF (column 2)
for orf_id, qgroup in groupby(blast_tab_fh,
key=lambda x: x.split("\t")[1]):
protein_function = "hypothetical protein"
protein_set = False
taxonomy_id = "1"
bitscore = "NA"
evalue = "NA"
orf_hits = BlastHits(max_hits=max_hits_per_orf,
top_fraction=top_fraction_of_hits)
lines = []
# iterate over blast hits per ORF
for hsp in qgroup:
# HSPs will now have contig in column 1
toks = hsp.strip().split("\t")
# remove extra column from toks
contig_name = toks.pop(0)
# convert toks to dictionary
toks = dict(zip(BLAST6, toks))
if (int(toks["length"]) < min_length
or float(toks["pident"]) < min_identity
or float(toks["evalue"]) > max_evalue):
continue
if min_bitscore and float(toks["bitscore"]) < min_bitscore:
# input is sorted by decreasing bitscore
break
cursor.execute(
'SELECT function, taxonomy FROM %s WHERE name="%s"' %
(table_name, toks["sseqid"]))
current_function, current_taxonomy = cursor.fetchone()
# update taxonomy based on pident; would be similar to 16S taxonomy assignments
# current_taxonomy = tree.climb_tree(current_taxonomy, float(toks["pident"]))
if summary_method == "best":
taxonomy_id = current_taxonomy
protein_function = current_function
bitscore = toks["bitscore"]
evalue = toks["evalue"]
break
# TODO implement bitscore ratio as a measure of alignment quality as a function of input sequence
orf_hits.add(current_taxonomy, toks["bitscore"])
toks["current_function"] = current_function
toks["current_taxonomy"] = current_taxonomy
lines.append(toks)
# summary method is majority and we have passing HSPs
if not summary_method == "best" and lines:
if summary_method == "majority":
taxonomy_id = orf_hits.majority()
for toks in lines:
if toks["current_taxonomy"] == taxonomy_id:
bitscore = toks["bitscore"]
evalue = toks["evalue"]
protein_function = toks["current_function"]
break
# summary method is 'lca'
else:
orf_hits.names.reverse()
taxonomy_id = tree.lca(orf_hits.names,
threshold=lca_threshold)
# grabbing best hit's bitscore and evalue
bitscore = lines[0]["bitscore"]
evalue = lines[0]["evalue"]
protein_function = lines[0]["current_function"]
if bitscore == "NA":
logging.critical(
"The summarized ID (%s) was not assigned metadata" %
taxonomy_id)
contigs[contig_name][orf_id] = (protein_function, taxonomy_id,
bitscore, evalue)
return contigs
def eggnog_parser(tsv, namemap, output, summary_method, min_identity, min_bitscore, min_length, max_evalue, top_fraction, max_hits, table_name):
"""Parse BLAST hits from EGGNOG.
The BLAST hits are assumed to be sorted by query with decreasing bitscores (best alignment first):
\b
sort -k1,1 -k12,12rn tsv > sorted_tsv
Expected columns in the EggNOG database:
\b
uniprot_ac
eggnog_ssid_b
eggnog_species_id
uniprot_id
ko_id
ko_level1_name
ko_level2_name
ko_level3_id
ko_level3_name
ko_gene_symbol
ko_product
ko_ec
Args:
tsv (str): blast hits file path in default tabular format
namemap (str): sqlite database file path
output (str): :py:class:`click.File` in write mode
summary_method (str): either 'majority' or 'best'; summary method for annotating ORFs; when majority and there is no majority, best is used
min_identity (int): minimum allowable percent ID of BLAST hit
min_bitscore (int): minimum allowable bitscore of BLAST hit; 0 disables
min_length (int): minimum allowable BLAST alignment length
max_evalue (float): maximum allowable e-value of BLAST hit
top_fraction (float): filters ORF BLAST hits before finding majority by only keep hits within this fraction, e.g. 0.98, of the highest bitscore
max_hits (int): maximum number of BLAST hits to consider when summarizing ORFs as a majority
table_name (str): table name within namemap database; expected columns are listed above
"""
logging.info("Parsing %s" % tsv)
if top_fraction == 1:
top_fraction = None
print("contig", "orf", "uniprot_ac", "eggnog_ssid_b", "eggnog_species_id", "uniprot_id",
"ko_id", "ko_level1_name", "ko_level2_name", "ko_level3_id", "ko_level3_name",
"ko_gene_symbol", "ko_product", "ko_ec", "%s_evalue" % table_name,
"%s_bitscore" % table_name, sep="\t", file=output)
with contextlib.closing(sqlite3.connect(namemap)) as conn, gzopen(tsv) as blast_tab_fh:
cursor = conn.cursor()
for query, qgroup in groupby(blast_tab_fh, key=lambda x: x.partition("\t")[0]):
contig_name, _, orf_idx = query.rpartition("_")
hit_id, evalue, bitscore = get_hit_from_blast_group(qgroup, max_hits, top_fraction, min_length, min_identity, max_evalue, min_bitscore, summary_method)
uniprot_ac = "NA"
eggnog_ssid_b = "NA"
eggnog_species_id = "NA"
uniprot_id = "NA"
ko_id = "NA"
ko_level1_name = "NA"
ko_level2_name = "NA"
ko_level3_id = "NA"
ko_level3_name = "NA"
ko_gene_symbol = "NA"
ko_product = "NA"
ko_ec = "NA"
# everything could have been filtered out due to user constraints
if hit_id:
cursor.execute('SELECT uniprot_ac, eggnog_ssid_b, eggnog_species_id, uniprot_id, \
ko_id, ko_level1_name, ko_level2_name, \
ko_level3_id, ko_level3_name, ko_gene_symbol, ko_product, ko_ec \
FROM %s \
WHERE eggnog_ssid_b="%s"' % (table_name, hit_id))
try:
uniprot_ac, eggnog_ssid_b, eggnog_species_id, uniprot_id, ko_id, \
ko_level1_name, ko_level2_name, ko_level3_id, ko_level3_name, \
ko_gene_symbol, ko_product, ko_ec = cursor.fetchone()
# legacy before database was pruned; can have hits not in metadata
except TypeError:
logging.warning("'%s' not present in database" % hit_id)
pass
# print for this query
print(contig_name, "%s_%s" % (contig_name, orf_idx), uniprot_ac, eggnog_ssid_b,
eggnog_species_id, uniprot_id, ko_id, ko_level1_name, ko_level2_name,
ko_level3_id, ko_level3_name, ko_gene_symbol, ko_product, ko_ec, evalue,
bitscore, sep="\t", file=output)
logging.info("Complete")