def testParseHmmOutput(self):
actual_hmmsearch = list(
hmmer_utils.parse_hmmer_output(_HMMER_TBLOUT, 'PF00131.19'))
print(actual_hmmsearch)
expected_hmmsearch = [
hmmer_utils.HMMEROutput(
sequence_name='MT4_CANLF/1-62',
predicted_label='PF00131.19',
true_label='PF00131.19',
score=60.9,
domain_evalue=1.3e-15,
),
hmmer_utils.HMMEROutput(
sequence_name='E4X7F8_OIKDI/453-561',
predicted_label='PF00131.19',
true_label='PF05033.15',
score=13.9,
domain_evalue=0.6,
),
]
self.assertEqual(actual_hmmsearch, expected_hmmsearch)
# Test that for the output file with no hits, the sentinel value is written.
actual_no_output = list(
hmmer_utils.parse_hmmer_output(_HMMER_TBLOUT_NO_OUTPUT,
'PF00131.19'))
print(actual_no_output)
expected_no_output = [
hmmer_utils.HMMEROutput(
sequence_name='no_sequence/0-0',
predicted_label='PF00131.19',
true_label='PF00000.0',
score=hmmer_utils.NO_SEQUENCE_MATCH_SCORE_SENTINEL,
domain_evalue=hmmer_utils.
NO_SEQUENCE_MATCH_DOMAIN_EVALUE_SENTINEL,
)
]
self.assertEqual(actual_no_output, expected_no_output)
def testFormatAsCsvHmmerOutput(self):
hmmer_output = hmmer_utils.HMMEROutput(
sequence_name='MT4_CANLF/1-62',
predicted_label='PF00131.19',
true_label='PF12345.6',
score=60.9,
domain_evalue=1.3e-15,
)
# order should match hmmer_utils.HMMER_OUTPUT_CSV_COLUMN_HEADERS.
# (That is, util.PREDICTION_FILE_COLUMN_NAMES + [DATAFRAME_SCORE_NAME_KEY].)
expected = 'MT4_CANLF/1-62,PF12345.6,PF00131.19,60.9,1.3e-15'
actual = hmmer_output.format_as_csv()
self.assertEqual(actual, expected)
class HMMerUtilsTest(parameterized.TestCase):
def testGetFamilyNameFromUnderscores(self):
# Sequence name contains underscores.
actual = hmmer_utils.get_family_name_from('V2R_HUMAN/54-325_PF00001.20')
expected = 'PF00001.20'
self.assertEqual(actual, expected)
def testGetFamilyNameFromNoUnderscores(self):
# Sequence name has no underscores.
actual = hmmer_utils.get_family_name_from('Q12345_alanine')
expected = 'alanine'
self.assertEqual(actual, expected)
def testGetSequenceNameFrom(self):
actual = hmmer_utils.get_sequence_name_from('V2R_HUMAN/54-325_PF00001.20')
expected = 'V2R_HUMAN/54-325'
self.assertEqual(actual, expected)
def testFormatAsCsvHmmerOutput(self):
hmmer_output = hmmer_utils.HMMEROutput(
sequence_name='MT4_CANLF/1-62',
predicted_label='PF00131.19',
true_label='PF12345.6',
score=60.9,
domain_evalue=1.3e-15,
)
# order should match hmmer_utils.HMMER_OUTPUT_CSV_COLUMN_HEADERS.
# (That is, util.PREDICTION_FILE_COLUMN_NAMES + [DATAFRAME_SCORE_NAME_KEY].)
expected = 'MT4_CANLF/1-62,PF12345.6,PF00131.19,60.9,1.3e-15'
actual = hmmer_output.format_as_csv()
self.assertEqual(actual, expected)
def testParseHmmOutput(self):
actual_hmmsearch = list(
hmmer_utils.parse_hmmer_output(_HMMER_TBLOUT, 'PF00131.19'))
print(actual_hmmsearch)
expected_hmmsearch = [
hmmer_utils.HMMEROutput(
sequence_name='MT4_CANLF/1-62',
predicted_label='PF00131.19',
true_label='PF00131.19',
score=60.9,
domain_evalue=1.3e-15,
),
hmmer_utils.HMMEROutput(
sequence_name='E4X7F8_OIKDI/453-561',
predicted_label='PF00131.19',
true_label='PF05033.15',
score=13.9,
domain_evalue=0.6,
),
]
self.assertEqual(actual_hmmsearch, expected_hmmsearch)
# Test that for the output file with no hits, the sentinel value is written.
actual_no_output = list(
hmmer_utils.parse_hmmer_output(_HMMER_TBLOUT_NO_OUTPUT, 'PF00131.19'))
print(actual_no_output)
expected_no_output = [
hmmer_utils.HMMEROutput(
sequence_name='no_sequence/0-0',
predicted_label='PF00131.19',
true_label='PF00000.0',
score=hmmer_utils.NO_SEQUENCE_MATCH_SCORE_SENTINEL,
domain_evalue=hmmer_utils.NO_SEQUENCE_MATCH_DOMAIN_EVALUE_SENTINEL,
)
]
self.assertEqual(actual_no_output, expected_no_output)
def testFilterFastaFileBySequenceName(self):
fasta_file_contents = ('>A0A0F7V1V9_TOXGV/243-280_PF10417.9\n'
'MIREVEKNGGKQVCPANWRRGEKMMHASFEGVKNYLGQ\n'
'>Q1QPP6_NITHX/169-202_PF10417.9\n'
'ALQATMSGQKLAPANWQPGETLLLPADEKTQKDT\n'
'>Q2K9A0_RHIEC/165-202_PF10417.9\n'
'SIQLTAKHQVATPANWNQGEDVIITAAVSNDDAIARFG\n')
input_fasta_file_name = test_util.tmpfile('input_fasta')
with tf.io.gfile.GFile(input_fasta_file_name, 'w') as input_fasta_file:
input_fasta_file.write(fasta_file_contents)
actual = list(
hmmer_utils.filter_fasta_file_by_sequence_name(
input_fasta_file_name, ['Q2K9A0_RHIEC/165-202']))
expected = [('>Q2K9A0_RHIEC/165-202_PF10417.9\n'
'SIQLTAKHQVATPANWNQGEDVIITAAVSNDDAIARFG\n')]
self.assertEqual(actual, expected)
def testAllSequenceNamesFromFastaFile(self):
fasta_file_contents = ('>A0A0F7V1V9_TOXGV/243-280_PF10417.9\n'
'MIREVEKNGGKQVCPANWRRGEKMMHASFEGVKNYLGQ\n'
'>Q1QPP6_NITHX/169-202_PF10417.9\n'
'ALQATMSGQKLAPANWQPGETLLLPADEKTQKDT\n'
'>Q2K9A0_RHIEC/165-202_PF10417.9\n'
'SIQLTAKHQVATPANWNQGEDVIITAAVSNDDAIARFG\n')
input_fasta_file_name = test_util.tmpfile('input_fasta')
with tf.io.gfile.GFile(input_fasta_file_name, 'w') as input_fasta_file:
input_fasta_file.write(fasta_file_contents)
actual = hmmer_utils.all_sequence_names_from_fasta_file(
input_fasta_file_name)
expected = [
'A0A0F7V1V9_TOXGV/243-280', 'Q1QPP6_NITHX/169-202',
'Q2K9A0_RHIEC/165-202'
]
np.testing.assert_array_equal(actual, expected)
def testSequencesWithNoPrediction(self):
hmmer_predictions = pd.DataFrame(
[['A4YXG4_BRASO/106-134', 'PF00001.1', 'PF00001.1', 0.5]],
columns=util.PREDICTION_FILE_COLUMN_NAMES)
all_sequence_names = ['A_DIFFERENTSEQNAME/1-2', 'A4YXG4_BRASO/106-134']
actual = hmmer_utils.sequences_with_no_prediction(
all_sequence_names=all_sequence_names,
hmmer_predictions=hmmer_predictions)
expected = {'A_DIFFERENTSEQNAME/1-2'}
self.assertEqual(actual, expected)
def testYieldTopElByScoreForEachSequenceName(self):
input_df = pd.DataFrame([
['SAME_SEQ_NAME', 'PF00264.20', 'PF000264.20', 10.1, 1e-3],
['SAME_SEQ_NAME', 'PF00264.20', 'PF00001.3', -65432.0, 100.],
],
columns=hmmer_utils.HMMER_OUTPUT_CSV_COLUMN_HEADERS)
actual = pd.concat(
list(
hmmer_utils.yield_top_el_by_score_for_each_sequence_name(input_df)))
expected = pd.DataFrame(
[['SAME_SEQ_NAME', 'PF00264.20', 'PF000264.20', 10.1, 1e-3]],
columns=hmmer_utils.HMMER_OUTPUT_CSV_COLUMN_HEADERS)
self.assertLen(actual, 1)
self.assertCountEqual(
actual.to_dict('records'), expected.to_dict('records'))
# pylint: disable=line-too-long
# Disable line-too-long because phmmer output strings are long, and we
# want to paste them verbatim.
@parameterized.named_parameters(
dict(
testcase_name=('multiple seq outputs, no repeats, same families, all '
'identifiers have predictions'),
phmmer_output="""# --- full sequence ---- --- best 1 domain ---- --- domain number estimation ----
# target name accession query name accession E-value score bias E-value score bias exp reg clu ov env dom rep inc description of target
# ------------------- ---------- -------------------- ---------- --------- ------ ----- --------- ------ ----- --- --- --- --- --- --- --- --- ---------------------
OPSB_HUMAN/51-303_PF00001.20 - OPSB_HUMAN/51-303_PF00001.20 - 2.5e-61 193.4 64.0 1.6e-22 66.2 0.9 8.7 8 1 0 8 8 8 8 -
OPS3_DROME/75-338_PF00001.20 - OPS3_DROME/75-338_PF00001.20 - 2.1e-69 219.9 65.6 3.2e-21 62.0 0.3 9.8 10 1 0 10 10 10 10 -
#
# Program: phmmer
# Version: 3.1b2 (February 2015)
# Pipeline mode: SEARCH
# Query file: /storage/hmm_train/PF00001.20.fasta
# Target file: /storage/hmm_train/PF00001.20.fasta
# Option settings: phmmer -o /dev/null --tblout /dev/stdout -E 10 /storage/hmm_train/PF00001.20.fasta /storage/hmm_train/PF00001.20.fasta
# Date: Wed Nov 21 09:26:50 2018
# [ok]
""",
all_identifiers=[
'OPSB_HUMAN/51-303_PF00001.20',
'OPS3_DROME/75-338_PF00001.20',
],
expected=[
hmmer_utils.HMMEROutput(
sequence_name='OPSB_HUMAN/51-303',
true_label='PF00001.20',
predicted_label='PF00001.20',
score=193.4,
domain_evalue=2.5e-61,
),
hmmer_utils.HMMEROutput(
sequence_name='OPS3_DROME/75-338',
true_label='PF00001.20',
predicted_label='PF00001.20',
score=219.9,
domain_evalue=2.1e-69,
)
],
),
dict(
testcase_name=('one seq output, no repeats, same families, some '
'identifiers do not have predictions'),
phmmer_output="""# --- full sequence ---- --- best 1 domain ---- --- domain number estimation ----
# target name accession query name accession E-value score bias E-value score bias exp reg clu ov env dom rep inc description of target
# ------------------- ---------- -------------------- ---------- --------- ------ ----- --------- ------ ----- --- --- --- --- --- --- --- --- ---------------------
OPSB_HUMAN/51-303_PF00001.20 - OPSB_HUMAN/51-303_PF00001.20 - 2.5e-61 193.4 64.0 1.6e-22 66.2 0.9 8.7 8 1 0 8 8 8 8 -
#
# Program: phmmer
# Version: 3.1b2 (February 2015)
# Pipeline mode: SEARCH
# Query file: /storage/hmm_train/PF00001.20.fasta
# Target file: /storage/hmm_train/PF00001.20.fasta
# Option settings: phmmer -o /dev/null --tblout /dev/stdout -E 10 /storage/hmm_train/PF00001.20.fasta /storage/hmm_train/PF00001.20.fasta
# Date: Wed Nov 21 09:26:50 2018
# [ok]
""",
expected=[
hmmer_utils.HMMEROutput(
sequence_name='OPSB_HUMAN/51-303',
true_label='PF00001.20',
predicted_label='PF00001.20',
score=193.4,
domain_evalue=2.5e-61,
),
hmmer_utils.HMMEROutput(
sequence_name='THIS_ISNOTSEEN/1-111',
true_label='PF09876.5',
predicted_label='PF00000.0',
score=hmmer_utils.NO_SEQUENCE_MATCH_SCORE_SENTINEL,
domain_evalue=hmmer_utils.NO_SEQUENCE_MATCH_DOMAIN_EVALUE_SENTINEL,
)
],
all_identifiers=[
'OPSB_HUMAN/51-303_PF00001.20',
'THIS_ISNOTSEEN/1-111_PF09876.5',
]),
dict(
testcase_name=('one seq output, with repeats, all identifiers have '
'predictions'),
phmmer_output="""# --- full sequence ---- --- best 1 domain ---- --- domain number estimation ----
# target name accession query name accession E-value score bias E-value score bias exp reg clu ov env dom rep inc description of target
# ------------------- ---------- -------------------- ---------- --------- ------ ----- --------- ------ ----- --- --- --- --- --- --- --- --- ---------------------
SSR1_HUMAN/75-323_PF00001.20 - CXCR5_HUMAN/68-322_PF00001.20 - 1e-06 14.4 0.9 1e-06 14.4 0.9 2.0 2 0 0 2 2 2 2 -
CX3C1_RAT/49-294_PF00001.20 - CXCR5_HUMAN/68-322_PF00001.20 - 1.3e-06 14.1 8.7 4.5e-06 12.3 1.7 2.6 3 0 0 3 3 3 2 -
#
# Program: phmmer
# Version: 3.1b2 (February 2015)
# Pipeline mode: SEARCH
# Query file: /storage/hmm_train/PF00001.20.fasta
# Target file: /storage/hmm_train/PF00001.20.fasta
# Option settings: phmmer -o /dev/null --tblout /dev/stdout -E 10 /storage/hmm_train/PF00001.20.fasta /storage/hmm_train/PF00001.20.fasta
# Date: Wed Nov 21 09:26:50 2018
# [ok]
""",
expected=[
hmmer_utils.HMMEROutput(
sequence_name='CXCR5_HUMAN/68-322',
true_label='PF00001.20',
predicted_label='PF00001.20',
score=14.4,
domain_evalue=1e-6,
),
hmmer_utils.HMMEROutput(
sequence_name='CXCR5_HUMAN/68-322',
true_label='PF00001.20',
predicted_label='PF00001.20',
score=14.1,
domain_evalue=1.3e-6,
)
],
all_identifiers=[
'CXCR5_HUMAN/68-322_PF00001.20',
]),
dict(
testcase_name=('no hits for any input sequence'),
phmmer_output="""
# --- full sequence ---- --- best 1 domain ---- --- domain number estimation ----
# target name accession query name accession E-value score bias E-value score bias exp reg clu ov env dom rep inc description of target
#------------------- ---------- -------------------- ---------- --------- ------ ----- --------- ------ ----- --- --- --- --- --- --- --- --- ---------------------
#
# Program: phmmer
# Version: 3.1b2 (February 2015)
# Pipeline mode: SEARCH
# Query file: -
# Target file: all_trainseqs.fasta
# Option settings: phmmer -o /dev/null --tblout /dev/stdout -E 10.0 - all_trainseqs.fasta
# Date: Mon Oct 29 12:45:57 2018
# [ok]
""",
all_identifiers=[
'CXCR5_HUMAN/68-322_PF00001.20',
],
expected=[
hmmer_utils.HMMEROutput(
sequence_name='CXCR5_HUMAN/68-322',
true_label='PF00001.20',
predicted_label=hmmer_utils.NO_SEQUENCE_MATCH_FAMILY_NAME_SENTINEL,
score=hmmer_utils.NO_SEQUENCE_MATCH_SCORE_SENTINEL,
domain_evalue=hmmer_utils.NO_SEQUENCE_MATCH_DOMAIN_EVALUE_SENTINEL,
),
],
),
)
def testParsePhmmer(self, phmmer_output, all_identifiers, expected):
actual = hmmer_utils.parse_phmmer_output(phmmer_output, all_identifiers)
self.assertEqual(actual, expected)