def set_background(self, fname=None, genome=None, length=200, nseq=10000):
"""Set the background to use for FPR and z-score calculations.
Background can be specified either as a genome name or as the
name of a FASTA file.
Parameters
----------
fname : str, optional
Name of FASTA file to use as background.
genome : str, optional
Name of genome to use to retrieve random sequences.
length : int, optional
Length of genomic sequences to retrieve. The default
is 200.
nseq : int, optional
Number of genomic sequences to retrieve.
"""
length = int(length)
if genome and fname:
raise ValueError("Need either genome or filename for background.")
if fname:
if not os.path.exists(fname):
raise IOError(
"Background file {} does not exist!".format(fname))
self.background = Fasta(fname)
self.background_hash = file_checksum(fname)
return
if not genome:
if self.genome:
genome = self.genome
logger.info(
"Using default background: genome {} with length {}".
format(genome, length))
else:
raise ValueError(
"Need either genome or filename for background.")
logger.info("Using background: genome {} with length {}".format(
genome, length))
with Cache(CACHE_DIR) as cache:
self.background_hash = "{}\{}".format(genome, int(length))
fa = cache.get(self.background_hash)
if not fa:
fa = RandomGenomicFasta(genome, length, nseq)
cache.set(self.background_hash, fa)
self.background = fa
def set_background(self,
fname=None,
genome=None,
size=200,
nseq=10000,
gc=False,
gc_bins=None):
"""Set the background to use for FPR and z-score calculations.
Background can be specified either as a genome name or as the
name of a FASTA file.
Parameters
----------
fname : str, optional
Name of FASTA file to use as background.
genome : str, optional
Name of genome to use to retrieve random sequences.
size : int, optional
Size of genomic sequences to retrieve. The default
is 200.
nseq : int, optional
Number of genomic sequences to retrieve.
"""
if self.background:
return
size = int(size)
if genome and fname:
raise ValueError("Need either genome or filename for background.")
if fname:
if not os.path.exists(fname):
raise IOError(
"Background file {} does not exist!".format(fname))
self.background = Fasta(fname)
self.background_hash = file_checksum(fname)
return
if not genome:
if self.genome:
genome = self.genome
else:
raise ValueError(
"Need either genome or filename for background.")
logger.debug("using background: genome {} with size {}".format(
genome, size))
lock.acquire()
with Cache(CACHE_DIR) as cache:
self.background_hash = "d{}:{}:{}:{}".format(
genome, int(size), gc, str(gc_bins))
c = cache.get(self.background_hash)
if c:
fa, gc_bins = c
else:
fa = None
if not fa:
if gc:
if gc_bins is None:
gc_bins = [(0.0, 0.2), (0.8, 1)]
for b in np.arange(0.2, 0.799, 0.05):
gc_bins.append((b, b + 0.05))
with NamedTemporaryFile() as tmp:
logger.info("using {} sequences".format(nseq))
gc_bin_bedfile(tmp.name,
genome,
number=nseq,
length=size,
bins=gc_bins)
fa = as_fasta(tmp.name, genome=genome)
else:
fa = RandomGenomicFasta(genome, size, nseq)
cache.set(self.background_hash, (fa, gc_bins))
lock.release()
self.background = fa
if gc_bins:
self.gc_bins = gc_bins
def set_threshold(self,
fpr=None,
threshold=None,
genome=None,
length=200,
filename=None):
"""Set motif scanning threshold based on background sequences.
Parameters
----------
fpr : float, optional
Desired FPR, between 0.0 and 1.0.
threshold : float or str, optional
Desired motif threshold, expressed as the fraction of the
difference between minimum and maximum score of the PWM.
Should either be a float between 0.0 and 1.0 or a filename
with thresholds as created by 'gimme threshold'.
"""
if threshold:
if fpr:
raise ValueError("Need either fpr or threshold.")
if genome:
sys.stderr.write(
"Parameter genome ignored when threshold is specified.\n"
"Did you want to use fpr?\n")
if filename:
sys.stderr.write(
"Parameter filename ignored when threshold is specified.\n"
"Did you want to use fpr?\n")
if genome and filename:
raise ValueError("Need either genome or filename.")
if fpr:
fpr = float(fpr)
if not (0.0 < fpr < 1.0):
raise ValueError("Parameter fpr should be between 0 and 1")
if not self.motifs:
raise ValueError("please run set_motifs() first")
thresholds = {}
with open(self.motifs) as f:
motifs = read_motifs(f)
if threshold is not None:
self.threshold = parse_threshold_values(self.motifs, threshold)
return
if filename:
if not os.path.exists(filename):
raise IOError("File {} does not exist.".format(filename))
bg_hash = file_checksum(filename)
seqs = Fasta(filename).seqs
elif genome:
bg_hash = "{}\{}".format(genome, int(length))
else:
raise ValueError("Need genome or filename")
with Cache(CACHE_DIR) as cache:
scan_motifs = []
for motif in motifs:
k = "{}|{}|{:.4f}".format(motif.hash(), bg_hash, fpr)
threshold = cache.get(k)
if threshold is None:
scan_motifs.append(motif)
else:
if np.isclose(threshold, motif.pwm_max_score()):
thresholds[motif.id] = None
else:
thresholds[motif.id] = threshold
if len(scan_motifs) > 0:
if genome:
Genome(genome)
sys.stderr.write(
"Determining threshold for fpr {} and length {} based on {}\n"
.format(fpr, int(length), genome))
fa = RandomGenomicFasta(genome, length, 10000)
seqs = fa.seqs
else:
sys.stderr.write(
"Determining threshold for fpr {} based on {}\n".
format(fpr, filename))
for motif, threshold in self._threshold_from_seqs(
scan_motifs, seqs, fpr):
k = "{}|{}|{:.4f}".format(motif.hash(), bg_hash, fpr)
cache.set(k, threshold)
if np.isclose(threshold, motif.pwm_max_score()):
thresholds[motif.id] = None
else:
thresholds[motif.id] = threshold
self.threshold_str = "{}_{}_{}_{}_{}".format(fpr, threshold, genome,
length, filename)
self.threshold = thresholds