def _get_pval_uncorr(self, study, log=sys.stdout):
"""Calculate the uncorrected pvalues for study items."""
log.write("Calculating uncorrected p-values using {PFNC}\n".format(PFNC=self.pval_obj.name))
results = []
go2studyitems = get_terms("study", study, self.assoc, self.obo_dag, log)
pop_n, study_n = self.pop_n, len(study)
allterms = set(go2studyitems.keys()).union(
set(self.go2popitems.keys()))
calc_pvalue = self.pval_obj.calc_pvalue
for term in allterms:
study_items = go2studyitems.get(term, set())
study_count = len(study_items)
pop_items = self.go2popitems.get(term, set())
pop_count = len(pop_items)
one_record = GOEnrichmentRecord(
GO=term,
p_uncorrected=calc_pvalue(study_count, study_n, pop_count, pop_n),
study_items=study_items,
pop_items=pop_items,
ratio_in_study=(study_count, study_n),
ratio_in_pop=(pop_count, pop_n))
results.append(one_record)
return results
def get_pval_uncorr(self, study, log=sys.stdout):
"""Calculate the uncorrected pvalues for study items."""
results = []
go2studyitems = get_terms("study", study, self.assoc, self.obo_dag, log)
pop_n, study_n = self.pop_n, len(study)
allterms = set(go2studyitems).union(set(self.go2popitems))
if log is not None:
log.write("Calculating {N:,} uncorrected p-values using {PFNC}\n".format(
N=len(allterms), PFNC=self.pval_obj.name))
calc_pvalue = self.pval_obj.calc_pvalue
for term in allterms:
study_items = go2studyitems.get(term, set())
study_count = len(study_items)
pop_items = self.go2popitems.get(term, set())
pop_count = len(pop_items)
one_record = GOEnrichmentRecord(
GO=term,
p_uncorrected=calc_pvalue(study_count, study_n, pop_count, pop_n),
study_items=study_items,
pop_items=pop_items,
ratio_in_study=(study_count, study_n),
ratio_in_pop=(pop_count, pop_n))
results.append(one_record)
return results
def __init__(self,
pop,
assoc,
obo_dag,
propagate_counts=True,
alpha=.05,
methods=None,
**kws):
self.log = kws['log'] if 'log' in kws else sys.stdout
self._run_multitest = {
'local': lambda iargs: self._run_multitest_local(iargs),
'statsmodels': lambda iargs: self._run_multitest_statsmodels(iargs)
}
self.pop = pop
self.pop_n = len(pop)
self.assoc = assoc
self.obo_dag = obo_dag
self.alpha = alpha
if methods is None:
methods = ["bonferroni", "sidak", "holm"]
self.methods = Methods(methods)
self.pval_obj = FisherFactory(**kws).pval_obj
if propagate_counts:
sys.stderr.write("Propagating term counts to parents ..\n")
obo_dag.update_association(assoc)
self.go2popitems = get_terms("population", pop, assoc, obo_dag,
self.log)
def __init__(self,
pop,
assoc,
obo_dag,
propagate_counts=True,
alpha=.05,
methods=None,
**kws):
self.name = kws.get('name', 'GOEA')
print('\nLoad {OBJNAME} Gene Ontology Analysis ...'.format(
OBJNAME=self.name))
self.log = kws['log'] if 'log' in kws else sys.stdout
self._run_multitest = {
'local': self._run_multitest_local,
'statsmodels': self._run_multitest_statsmodels
}
self.pop = set(pop)
self.pop_n = len(pop)
self.assoc = assoc
self.obo_dag = obo_dag
self.alpha = alpha
if methods is None:
methods = ["bonferroni", "sidak", "holm"]
self.methods = Methods(methods)
self.pval_obj = FisherFactory(**kws).pval_obj
if propagate_counts:
update_association(assoc, obo_dag, kws.get('relationships', None))
## BROAD broad_goids = get_goids_to_remove(kws.get('remove_goids'))
## BROAD if broad_goids:
## BROAD assoc = self._remove_assc_goids(assoc, broad_goids)
self.go2popitems = get_terms("population", pop, assoc, obo_dag,
self.log)
def get_pval_uncorr(self, study, log=sys.stdout):
"""Calculate the uncorrected pvalues for study items."""
results = []
study_in_pop = self.pop.intersection(study)
# " 99% 378 of 382 study items found in population"
go2studyitems = get_terms("study", study_in_pop, self.assoc, self.obo_dag, log)
pop_n, study_n = self.pop_n, len(study_in_pop)
allterms = set(go2studyitems).union(set(self.go2popitems))
if log is not None:
study_n_orig = len(study)
log.write("{R:3.0f}% {N:>6,} of {M:>6,} study items found in population({P})\n".format(
N=study_n, M=study_n_orig, P=pop_n, R=100.0*study_n/study_n_orig))
log.write("Calculating {N:,} uncorrected p-values using {PFNC}\n".format(
N=len(allterms), PFNC=self.pval_obj.name))
calc_pvalue = self.pval_obj.calc_pvalue
for goid in allterms:
study_items = go2studyitems.get(goid, set())
study_count = len(study_items)
pop_items = self.go2popitems.get(goid, set())
pop_count = len(pop_items)
one_record = GOEnrichmentRecord(
GO=goid,
p_uncorrected=calc_pvalue(study_count, study_n, pop_count, pop_n),
study_items=study_items,
pop_items=pop_items,
ratio_in_study=(study_count, study_n),
ratio_in_pop=(pop_count, pop_n))
results.append(one_record)
return results
def __init__(self, pop, assoc, obo_dag, propagate_counts=True, alpha=.05, methods=None, **kws):
self.log = kws['log'] if 'log' in kws else sys.stdout
self._run_multitest = {
'local':lambda iargs: self._run_multitest_local(iargs),
'statsmodels':lambda iargs: self._run_multitest_statsmodels(iargs)}
self.pop = pop
self.pop_n = len(pop)
self.assoc = assoc
self.obo_dag = obo_dag
self.alpha = alpha
if methods is None:
methods = ["bonferroni", "sidak", "holm"]
self.methods = Methods(methods)
self.pval_obj = FisherFactory(**kws).pval_obj
if propagate_counts:
sys.stderr.write("Propagating term counts to parents ..\n")
obo_dag.update_association(assoc)
self.go2popitems = get_terms("population", pop, assoc, obo_dag, self.log)
def get_pval_uncorr(self, study, log=sys.stdout):
"""Calculate the uncorrected pvalues for study items."""
results = []
study_in_pop = self.pop.intersection(study)
# " 99% 378 of 382 study items found in population"
go2studyitems = get_terms("study", study_in_pop, self.assoc,
self.obo_dag, log)
pop_n, study_n = self.pop_n, len(study_in_pop)
allterms = set(go2studyitems).union(set(self.go2popitems))
if log is not None:
# Some study genes may not have been found in the population. Report from orig
study_n_orig = len(study)
perc = 100.0 * study_n / study_n_orig if study_n_orig != 0 else 0.0
log.write(
"{R:3.0f}% {N:>6,} of {M:>6,} study items found in population({P})\n"
.format(N=study_n, M=study_n_orig, P=pop_n, R=perc))
if study_n:
log.write(
"Calculating {N:,} uncorrected p-values using {PFNC}\n".
format(N=len(allterms), PFNC=self.pval_obj.name))
# If no study genes were found in the population, return empty GOEA results
if not study_n:
return []
calc_pvalue = self.pval_obj.calc_pvalue
for goid in allterms:
study_items = go2studyitems.get(goid, set())
study_count = len(study_items)
pop_items = self.go2popitems.get(goid, set())
pop_count = len(pop_items)
one_record = GOEnrichmentRecord(
goid,
p_uncorrected=calc_pvalue(study_count, study_n, pop_count,
pop_n),
study_items=study_items,
pop_items=pop_items,
ratio_in_study=(study_count, study_n),
ratio_in_pop=(pop_count, pop_n))
results.append(one_record)
return results
def get_pval_uncorr(self, study, log=sys.stdout):
"""Calculate the uncorrected pvalues for study items."""
results = []
study_in_pop = self.pop.intersection(study)
# " 99% 378 of 382 study items found in population"
go2studyitems = get_terms("study", study_in_pop, self.assoc, self.obo_dag, log)
pop_n, study_n = self.pop_n, len(study_in_pop)
allterms = set(go2studyitems).union(set(self.go2popitems))
if log is not None:
# Some study genes may not have been found in the population. Report from orig
study_n_orig = len(study)
perc = 100.0*study_n/study_n_orig if study_n_orig != 0 else 0.0
log.write("{R:3.0f}% {N:>6,} of {M:>6,} study items found in population({P})\n".format(
N=study_n, M=study_n_orig, P=pop_n, R=perc))
if study_n:
log.write("Calculating {N:,} uncorrected p-values using {PFNC}\n".format(
N=len(allterms), PFNC=self.pval_obj.name))
# If no study genes were found in the population, return empty GOEA results
if not study_n:
return []
calc_pvalue = self.pval_obj.calc_pvalue
for goid in allterms:
study_items = go2studyitems.get(goid, set())
study_count = len(study_items)
pop_items = self.go2popitems.get(goid, set())
pop_count = len(pop_items)
one_record = GOEnrichmentRecord(
goid,
p_uncorrected=calc_pvalue(study_count, study_n, pop_count, pop_n),
study_items=study_items,
pop_items=pop_items,
ratio_in_study=(study_count, study_n),
ratio_in_pop=(pop_count, pop_n))
results.append(one_record)
return results
def _get_pval_uncorr(self, study, log=sys.stdout):
"""Calculate the uncorrected pvalues for study items."""
if log is not None:
log.write("Calculating uncorrected p-values using {PFNC}\n".format(
PFNC=self.pval_obj.name))
go2studyitems = get_terms("study", study, self.assoc, self.obo_dag,
log)
pop_n, study_n = self.pop_n, len(study)
allterms = set(go2studyitems.keys()).union(set(
self.go2popitems.keys()))
# if self.pval_obj.log is a file handle, which we can not serialize, so we could
# not transfer self.pval_obj.calc_pvalue to another python process with multiprocessing.
# there fore we "path" the object which will later be restored again.
old = self.pval_obj.log
self.pval_obj.log = None
calc_pvalue = self.pval_obj.calc_pvalue
# -1 avoids freezing of the machine:
n_procs = multiprocessing.cpu_count() - 1
p = multiprocessing.Pool(n_procs)
n = len(allterms)
allterms = list(allterms)
fragments = [allterms[i::n_procs] for i in range(n_procs)]
remote_func = partial(compute_pvals,
calc_pvalue=calc_pvalue,
go2studyitems=go2studyitems,
go2popitems=self.go2popitems,
study_n=study_n,
pop_n=pop_n)
all_p_values = p.map(remote_func, fragments)
# restore patched file handle
self.pval_obj.log = old
results = []
for p_values in all_p_values:
for term, p_value in p_values.items():
study_items = go2studyitems.get(term, set())
study_count = len(study_items)
pop_items = self.go2popitems.get(term, set())
pop_count = len(pop_items)
one_record = GOEnrichmentRecord(GO=term,
p_uncorrected=p_value,
study_items=study_items,
pop_items=pop_items,
ratio_in_study=(study_count,
study_n),
ratio_in_pop=(pop_count,
pop_n))
results.append(one_record)
return results