def compute_silent_contamination(fna, faa, hits):
prot_contigs = set()
for seq in Fasta(faa):
contig = seq.name.split("_", 1)[1].rsplit("_", 1)[0]
prot_contigs.add(contig)
hit_contigs = set()
for row in hits:
contig = row["target"].split("_", 1)[1].rsplit("_", 1)[0]
hit_contigs.add(contig)
contigs = {}
total_bp = 0
for rec in Fasta(fna):
contigs[rec.name] = len(rec.seq)
total_bp += len(rec.seq)
# count all the stats
stats = {
"contigs": len(contigs),
"contigs_w_hits": 0,
"bp_w_hits": 0,
"contigs_w_proteins": 0,
"total_bp": total_bp,
}
for contig, size in contigs.items():
if contig in prot_contigs:
stats["contigs_w_proteins"] += 1
if contig in hit_contigs:
stats["contigs_w_hits"] += 1
stats["bp_w_hits"] += size
return stats
def test_metaeuk_cleaner(self):
cleaned = Fasta(
clean_metaeuk_fasta(METAEUK_dirty, ".test_tmp_metaeuk_file"))
clean = Fasta(METAEUK_clean)
for seq_1, seq_2 in zip(clean, cleaned):
self.assertEqual(seq_1.name, seq_2.name)
self.assertEqual(seq_1.seq, seq_2.seq)
os.remove(".test_tmp_metaeuk_file")
def test_open_invalid(self):
with self.assertRaises(ValueError):
x = Fasta(NOT_A_FASTA)
for s in x:
print(s)
with self.assertRaises(ValueError):
x = Fasta(CORRUPT_FILE)
for s in x:
print(s)
def test_Fasta_extract(self):
fa = Fasta(TESTDATA_FASTA)
for seq in fa:
if seq.name == "TEST":
break
outfile = ".test_fasta"
reduce_fasta(TESTDATA_FASTA, outfile, ["TEST"])
fa2 = Fasta(outfile)
for seq2 in fa2:
if seq2.name == "TEST":
break
self.assertEqual(seq.seq, seq2.seq)
self.assertEqual(seq.name, seq2.name)
self.assertEqual(seq.long_name, seq2.long_name)
os.remove(outfile)
def split_contig_faa(path, workdir, delim="_binsep_"):
faa_dir = os.path.join(workdir, "faa")
if os.path.exists(faa_dir):
logging.info(
"Faa folder exists, remove if you want to rerun this analsys, will reuse it for now"
)
return [
os.path.abspath(os.path.join(faa_dir, x))
for x in os.listdir(faa_dir)
]
elif file.isdir(faa_dir):
fls = {}
for seq in Fasta(path):
b = seq.name.split(delim, 1)[0]
contig = seq.name.split(delim, 1)[1]
if b not in fls.keys():
fls[b] = open(os.path.join(faa_dir, b), "w")
fls[b].write(">{name}\n{seq}\n".format(name=contig, seq=seq.seq))
# close all files
for key, fl in fls.items():
fl.close()
return [
os.path.abspath(os.path.join(faa_dir, x))
for x in os.listdir(faa_dir)
]
else:
logging.warning("Could not create fasta split dir: {}".format(faa_dir))
def read_fasta(path):
seqs = {}
for record in Fasta(path):
seqs[record.name] = record.seq
return seqs
def test_multiline(self):
expected = ["AAGGCCTT", "CCTTAAGG", "AAGGCCTT"]
for e, seq in zip(expected, Fasta(MULTILINE_FILE)):
self.assertEqual(e, seq.seq)
def test_Fasta(self):
fa = Fasta(TESTDATA_FASTA)
for seq in fa:
if seq.name == "TEST":
self.assertEqual(seq.seq, "HELLO")