def pseudo_bg(bgs, no_reverse):
"""
Add the pseudocount to the background frequencies
----
Parameters:
bgs (dict) : dictionary of the background frequencies
no_reverse (bool) : if set to True, the background
frequencies will be averaged with the
reverse complement frequencies
----
Returns:
bgs_proc (dict) : normalized (and averaged) background frequencies
"""
if not isinstance(bgs, dict):
raise NotValidBGException(
"\n\nERROR: unable to add the pseudocount to the background")
if not isinstance(no_reverse, bool):
raise ValueException(' '.join(
["Boolean value required, got",
str(type(no_reverse))]))
if not no_reverse:
bgs_avg = average_bg_with_rc(bgs)
else:
bgs_avg = bgs
bgs_proc = norm_bg(bgs_avg)
return bgs_proc
def print_scoring_msg(no_reverse, motif):
"""Print a message to display on terminal during scoring step of
GRAFIMO analysis.
Parameters
----------
no_reverse : bool
if True will be considered only the forward DNA strand
motif : Motif
Motif object
"""
if not isinstance(motif, Motif):
errmsg: str = '\n\nERROR: The given motif is not an instance of Motif'
raise ValueException(errmsg)
motif_id: str = motif.getMotifID()
fw_id: str = ''.join(['+', motif_id])
# we take into account also the reverse complement
if not no_reverse:
rev_id: str = ''.join(['-', motif_id])
print('\nScoring hits for motif', fw_id)
# if we score also the reverse complement
if not no_reverse:
print('Scoring hits for motif', rev_id, end="\n\n")
def compute_qvalues(pvalues: List[np.double]) -> List[np.double]:
"""Compute q-values for a given list of P-values.
The q-values are obtained using the Benjamini-Hochberg method.
Parameters
----------
pvalues : list
list of P-values
Returns
-------
list
list of q-values
"""
if not isinstance(pvalues, list):
errmsg: str = "\n\nERROR: P-values must be in a list"
raise ValueException(errmsg)
print("\nComputing q-values...\n")
# use Benjamini-Hochberg procedure to correct P-values
mt_obj: Tuple[np.ndarray, np.ndarray, np.double, float]
mt_obj = multipletests(pvalues, method="fdr_bh")
qvalues: List[float] = list(mt_obj[1])
return qvalues
def setScale(self, scale: int) -> None:
if not isinstance(scale, int):
raise ValueException("\n\nERROR: the scale factor must be an int")
assert scale > 0
self._scale = scale
def setMax_val(self, max_val: int) -> None:
if max_val >= np.inf:
errmsg = ' '.join([
"\n\nERROR: impossible to assign", max_val, "to Motif.max_val"
])
raise ValueException(errmsg)
self._max_val = max_val
def setMin_val(self, min_val):
if min_val <= -np.inf:
errmsg = ' '.join([
"\n\nERROR: impossible to assign", min_val, "to Motif.min_val"
])
raise ValueException(errmsg)
self._min_val = min_val
def pseudo_bg(bgs: Dict, no_reverse: bool) -> Dict:
"""Add a pseudocount and normalize the background probabilities of
nucleotides used to build the motif scoring matrix.
A pseudocount value is added to the background probability
distribution.
If are to be considered both the forward and the reverse strand the
background probabilities are averaged for the two strands.
The resulting background probabilities are then normalized.
Parameters
----------
bgs : dict
background probability distribution
no_reverse : bool
if False only the forward strand will be considered, otherwise
both forward and reverse are considered
Returns
-------
dict
normalized background probablity distribution
"""
bgs_avg: Dict
bgs_proc: Dict
errmsg: str
if not isinstance(bgs, dict):
errmsg = "\n\nERROR: unable to add the pseudocount to the background"
raise NotValidBGException(errmsg)
if not isinstance(no_reverse, bool):
errmsg = ' '.join(
["Boolean value required, got",
str(type(no_reverse))])
raise ValueException(errmsg)
if not no_reverse:
bgs_avg = average_bg_with_rc(bgs)
else:
bgs_avg = bgs
bgs_proc = norm_bg(bgs_avg)
return bgs_proc
def print_scoring_msg(no_reverse, motif):
if not isinstance(motif, Motif):
raise ValueException(
'\n\nERROR: The given motif is not an instance of Motif')
motif_id = motif.getMotifID()
fw_id = ''.join(['+', motif_id])
# we take into account also the reverse complement
if not no_reverse:
rev_id = ''.join(['-', motif_id])
print() # newline
print('Scoring hits for motif', fw_id)
# if we score also the reverse complement
if not no_reverse:
print('Scoring hits for motif', rev_id)
print() # newline
def compute_qvalues(pvalues):
"""
Compute the q-values for a given list
of P-values, using the Benjamini-Hochberg method
----
Parameters:
pvalues (list) : list of P-values
----
Returns:
qvalues (list) : list of computed q-values
"""
if not isinstance(pvalues, list):
errmsg = "\n\nERROR: P-values must be in a list"
raise ValueException(errmsg)
print("\nComputing q-values...\n")
# use Benjamini-Hochberg procedure to correct P-values
mt_obj = multipletests(pvalues, method="fdr_bh")
qvalues = list(mt_obj[1])
return qvalues
def build_df(motif: Motif,
seqnames: List[str],
starts: List[int],
stops: List[int],
strands: List[str],
scores: List[np.double],
pvalues: List[np.double],
qvalues: List[np.double],
sequences: List[str],
frequencies: List[int],
references: List[str],
threshold: float,
qval_t: bool,
no_qvalue: bool,
recomb: bool
) -> pd.DataFrame:
"""Build the results summary report. The results are stored in a
pandas DataFrame object.
The motif occurrence candidates are filtered applying a threshold on
the P-value or on the q-value.
The remaining entries are reported in the final results.
Parameters
----------
motif : Motif
Motif object
seqnames : list
sequence names
starts : list
starting coordinates
stops : list
stopping coordinates
strands : list
DNA strands
pvalues: list
P-values
qvalues : list
q-values
sequences : list
sequences
references : list
flag values stating if the sequences contain genomi variants
threshold : float
threshold to apply on P-values or q-values in order to filter
the motif occurrence candidates to report
qval_t : bool
if True the threshold will be applied on q-values rather on
P-values
no_qvalue:
if True the q-values have not been computed
recomb : bool
if True will be reported also sequences which can be built with
the given set of genomic variants but do not appear in the
available samples haplotypes
Returns
-------
pandas.DataFrame
final results report
"""
errmsg: str = "\n\nERROR: unknown data-type for motif"
if not isinstance(motif, Motif):
raise ValueException(errmsg)
if not isinstance(seqnames, list):
raise ValueException(errmsg)
if not isinstance(starts, list):
raise ValueException(errmsg)
if not isinstance(stops, list):
raise ValueException(errmsg)
if not isinstance(strands, list):
raise ValueException(errmsg)
if not isinstance(pvalues, list):
raise ValueException(errmsg)
if not isinstance(qvalues, list):
raise ValueException(errmsg)
if not isinstance(sequences, list):
raise ValueException(errmsg)
if not isinstance(references, list):
raise ValueException(errmsg)
if not isinstance(references, list):
raise ValueException(errmsg)
if not isinstance(qval_t, bool):
raise ValueException(errmsg)
if not isinstance(no_qvalue, bool):
raise ValueException(errmsg)
if not isinstance(recomb, bool):
raise ValueException(errmsg)
lst_len: int = len(seqnames)
assert len(starts) == lst_len
assert len(stops) == lst_len
assert len(strands) == lst_len
assert len(scores) == lst_len
assert len(pvalues) == lst_len
assert len(sequences) == lst_len
assert len(frequencies) == lst_len
assert len(references) == lst_len
# check if we want also the q-values
if not no_qvalue:
assert len(qvalues) == lst_len
# apply the threshold on the q-values rather than on P-values
if qval_t:
assert (not no_qvalue)
assert len(qvalues) > 0
seqnames_thresh: List[str] = list()
starts_thresh: List[int] = list()
ends_thresh: List[int] = list()
strands_thresh: List[str] = list()
scores_thresh: List[np.double] = list()
pvalues_thresh: List[np.double] = list()
sequences_thresh: List[str] = list()
frequencies_thresh: List[int] = list()
references_thresh: List[str] = list()
if not no_qvalue:
qvalues_thresh: List[np.double] = list()
for i in range(lst_len):
# ignore binding site candidates which does not appear in any sample
# if not required by tyhe user to analyze them
if not recomb and int(frequencies[i]) == 0:
continue
if not qval_t: # apply threshold on P-values
pvalue: np.double = pvalues[i]
if pvalue < threshold:
# only the sequences with a P-value under the threshold survive
seqnames_thresh.append(seqnames[i])
starts_thresh.append(starts[i])
ends_thresh.append(stops[i])
strands_thresh.append(strands[i])
scores_thresh.append(scores[i])
pvalues_thresh.append(pvalues[i])
sequences_thresh.append(sequences[i])
frequencies_thresh.append(frequencies[i])
references_thresh.append(references[i])
if not no_qvalue:
qvalues_thresh.append(qvalues[i])
# end if
else: # apply threshold on q-values
qvalue: np.double = qvalues[i]
if qvalue < threshold:
# only the sequences with a q-value under the threshold survive
seqnames_thresh.append(seqnames[i])
starts_thresh.append(starts[i])
ends_thresh.append(stops[i])
strands_thresh.append(strands[i])
scores_thresh.append(scores[i])
pvalues_thresh.append(pvalues[i])
sequences_thresh.append(sequences[i])
frequencies_thresh.append(frequencies[i])
references_thresh.append(references[i])
# the last control statement, in the if, in this case is not
# necessary (we must have the q-values)
# otherwise we should not be here
qvalues_thresh.append(qvalues[i])
# end if
# end if
# end for
df_len: int = len(seqnames_thresh)
# TF's name and ID list
motif_ids: List[str] = [motif.getMotifID()] * df_len
motif_names: List[str] = [motif.getMotifName()] * df_len
df = pd.DataFrame()
df['motif_id'] = motif_ids
df['motif_alt_id'] = motif_names
df['sequence_name'] = seqnames_thresh
df['start'] = starts_thresh
df['stop'] = ends_thresh
df['strand'] = strands_thresh
df['score'] = scores_thresh
df['p-value'] = pvalues_thresh
# add the q-values to the final data frame if they have been computed
if not no_qvalue:
df['q-value'] = qvalues_thresh
# finish to build the data frame
df['matched_sequence'] = sequences_thresh
df['haplotype_frequency'] = frequencies_thresh
df['reference'] = references_thresh
# sort entries by p-value
df = df.sort_values(['p-value'], ascending=True)
# reindex the data frame in order to have indexes in range [1, (df_len + 1)]
df.index = list(range(1, (df_len + 1)))
return df
def get_references(self):
if not self._references:
errmsg: str = "\n\nERROR: attempting to access an empty attribute"
raise ValueException(errmsg)
return self._references
def build_df(motif, seqnames, starts, stops, strands, scores, pvalues, qvalues,
sequences, references, threshold, qval_t, no_qvalue):
"""
Build a pandas DataFrame to summarize the results
of GRAFIMO analysis
----
Parameters:
motif (Motif) : motif
seqnames (list) : list of sequence names
starts (list) : list of sequence starting positions
stops (list) : list of sequence ending positions
strands (list) : list of sequence strands
scores (list) : list of sequence scores
pvalues (list) : list of sequence score P-values
qvalues (list) : list of sequence q-values
sequences (list) : list of sequences
references (list) : list of sequence flag values. If 'ref',
then the sequence belong to the reference genome,
if 'non.ref', then the sequence contains variants
threshold (float) : threshold to apply on the P-value (default behavior)
or on the q-values
qval_t (bool) : if set to True, the threshold will be applied on the
q-values, on the P-values otherwise
----
Returns:
df (pd.DataFrame)
"""
if not isinstance(motif, Motif):
errmsg = "\n\nERROR: unknown data-type for motif"
raise ValueException(errmsg)
if not isinstance(seqnames, list):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(starts, list):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(stops, list):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(strands, list):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(pvalues, list):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(qvalues, list):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(sequences, list):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(references, list):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(qval_t, bool):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
if not isinstance(no_qvalue, bool):
errmsg = "\n\nERROR: unknown data-type, cannot proceed"
raise ValueException(errmsg)
# all lists must have the same length
lst_len = len(seqnames)
assert len(starts) == lst_len
assert len(stops) == lst_len
assert len(strands) == lst_len
assert len(scores) == lst_len
assert len(pvalues) == lst_len
assert len(sequences) == lst_len
assert len(references) == lst_len
# check if we want also the q-values
if not no_qvalue: # we want the q-values
assert len(qvalues) == lst_len
if qval_t: # apply the threshold on the q-values rather than on P-values
assert (not no_qvalue)
assert len(qvalues) > 0 # we must have computed them
seqnames_thresh = []
starts_thresh = []
ends_thresh = []
strands_thresh = []
scores_thresh = []
pvalues_thresh = []
sequences_thresh = []
references_thresh = []
if not no_qvalue:
qvalues_thresh = []
for i in range(lst_len):
if not qval_t: # apply threshold on P-values
pvalue = pvalues[i]
if pvalue < threshold:
# only the sequences with a P-value under the threshold survive
seqnames_thresh.append(seqnames[i])
starts_thresh.append(starts[i])
ends_thresh.append(stops[i])
strands_thresh.append(strands[i])
scores_thresh.append(scores[i])
pvalues_thresh.append(pvalues[i])
sequences_thresh.append(sequences[i])
references_thresh.append(references[i])
if not no_qvalue:
qvalues_thresh.append(qvalues[i])
# end if
else: # apply threshold on q-values
qvalue = qvalues[i]
if qvalue < threshold:
# only the sequences with a q-value under the threshold survive
seqnames_thresh.append(seqnames[i])
starts_thresh.append(starts[i])
ends_thresh.append(stops[i])
strands_thresh.append(strands[i])
scores_thresh.append(scores[i])
pvalues_thresh.append(pvalues[i])
sequences_thresh.append(sequences[i])
references_thresh.append(references[i])
# the last control statement, in the if, in this case is not
# necessary (we must have the q-values)
# otherwise we should not be here
qvalues_thresh.append(qvalues[i])
# end if
# end if
# end for
df_len = len(seqnames_thresh)
# TF's name and ID list
motif_ids = [motif.getMotifID()] * df_len
motif_names = [motif.getMotifName()] * df_len
"""
build the final data frame
structure:
|motif_id|motif_alt_id|sequence_name|start|stop|strand|score|p-value|q-value|matched_sequence|reference|
"""
df = pd.DataFrame()
df['motif_id'] = motif_ids
df['motif_alt_id'] = motif_names
df['sequence_name'] = seqnames_thresh
df['start'] = starts_thresh
df['stop'] = ends_thresh
df['strand'] = strands_thresh
df['score'] = scores_thresh
df['p-value'] = pvalues_thresh
# add the q-values to the final data frame if they have been computed
if not no_qvalue:
df['q-value'] = qvalues_thresh
# finish to build the data frame
df['matched_sequence'] = sequences_thresh
df['reference'] = references_thresh
# sort entries by p-value
df = df.sort_values(['p-value'], ascending=True)
# reindex the data frame in order to have indexes in range [1, (df_len + 1)]
df.index = list(range(1, (df_len + 1)))
return df
def get_strands(self):
if not self._strands:
errmsg = "\n\nERROR: attempting to access an empty attribute"
raise ValueException(errmsg)
return self._strands
