def fit_dea(data, default_contrast, **kwargs):
# Create and fit analysis object
dea = DEAnalysis(data, **kwargs)
dea.fit_contrasts(dea.default_contrasts[default_contrast]['contrasts'],
fit_names=default_contrast)
der = dea.results[default_contrast] # type: DEResults
scores = der.score_clustering()
return dea, scores
def fit_dea(dea, data=None, override=False, **kwargs):
"""
:param data:
:param default_contrast:
:param kwargs:
:return:
"""
if dea is None or override:
# Set defaults
kwargs.setdefault('reference_labels', ['condition', 'time'])
kwargs.setdefault('index_names',
['condition', 'replicate', 'time'])
# Make the dea object and fit it
new_dea = DEAnalysis(data, **kwargs)
new_dea.fit_contrasts(new_dea.default_contrasts, status=True)
else:
new_dea = dea
return new_dea
def _read_permute(p, permutes_path, grouped):
# Load data and fit to get permuted data p-values
p_idx = p.split("_")[0]
p_dea = DEAnalysis(pd.read_pickle("{}/{}".format(permutes_path, p)),
time='Time',
replicate='rep',
reference_labels=['condition', 'Time'],
log2=False)
p_dea.fit_contrasts()
# Score ranking against the original clusters
p_der = p_dea.results['ko-wt']
weighted_lfc = (1 - p_der.p_value) * p_der.continuous.loc[
p_der.p_value.index, p_der.p_value.columns]
scores = get_scores(grouped, p_der.continuous.loc[:,
p_der.p_value.columns],
weighted_lfc).sort_index()
scores['score'] = scores['score'] * (
1 - p_der.continuous['adj_pval']).sort_index().values
scores.sort_values('score', ascending=False, inplace=True)
# drop cluster column, rename score, and add to real scores
scores.drop('Cluster', inplace=True, axis=1)
scores.columns = ['p{}_score'.format(p_idx)]
return scores
def load_sim_data(compiled_path):
# todo: needs to be fully functionalized
"""
Load simulation data to match
:param compiled_path:
:return:
"""
idx = pd.IndexSlice
# Load sim data and cleanup appropriately
sim_data = pd.read_pickle(compiled_path)
sim_data = sim_data.loc['y', idx[:, :, 1, :]]
sim_data.columns = sim_data.columns.remove_unused_levels()
sim_data.columns.set_names(['replicate', 'time'],
level=[1, 3],
inplace=True)
sim_data.columns.set_levels(['ki', 'pten', 'wt'], level=0, inplace=True)
sim_dea = DEAnalysis(sim_data,
reference_labels=contrast_labels,
index_names=sample_features)
return sim_dea
def fit_dea(path, data=None, override=False, **dea_kwargs):
"""
:param path:
:param override:
:return:
"""
dea_kwargs.setdefault('counts', True)
dea_kwargs.setdefault('log2', False)
try:
if override:
raise ValueError('Override to retrain')
dea = pd.read_pickle(path)
except (FileNotFoundError, ValueError) as e:
dea = DEAnalysis(data, **dea_kwargs)
dea.fit_contrasts(dea.default_contrasts)
dea.to_pickle(path)
return dea
get_data(ts_file,
input_type,
c,
n_timeseries,
reps,
p_labels,
t=t)
])
idx = pd.IndexSlice
data.sort_index(inplace=True)
# Censored object
dea = DEAnalysis(data.loc[idx['activating', ['wt', 'ko'], :,
perturb, :]].T,
time='Time',
replicate='rep',
reference_labels=['condition', 'Time'],
log2=False)
# Fit contrasts and save to pickle
if fit_data:
run_fit(dea, 'intermediate_data/{}_dea.pkl'.format(base_name))
# Save permuted data
if save_p:
grouped = dea.data.groupby(level='condition', axis=1)
save_permutes("intermediate_data/{}_permutes/".format(base_name),
grouped,
n=n_p)
if analyze_p:
import pandas as pd
from pydiffexp import DEAnalysis
# Load the data
data_path = "/Volumes/Hephaestus/jfinkle/Documents/Northwestern/MoDyLS/Code/Python/pydiffexp/data/GSE13100/" \
"bgcorrected_GSE13100_TR_data.pkl"
raw_data = pd.read_pickle(data_path)
# The example data has been background corrected, so set everything below 0 to a trivial positive value of 1
raw_data[raw_data <= 0] = 1
# Make the Differential Expression Analysis Object
# The reference labels specify how samples will be organized into unique values
dea = DEAnalysis(raw_data, replicate='rep', reference_labels=['condition', 'time'])
# Data can be standarized if desired
# norm_data = dea.standardize()
# Fit the contrasts and save the object
dea.fit_contrasts()
dea.to_pickle('intermediate_data/yeast_osmoTR_dea.pkl')
# Load the data
test_path = "/Users/jfinkle/Documents/Northwestern/MoDyLS/Code/Python/sprouty/data/raw_data/GSE63497_Oncogene_Formatted.tsv"
# test_path = "/Users/jfinkle/Documents/Northwestern/MoDyLS/Code/Python/sprouty/data/raw_data/GSE63497_VEC_CRE_Formatted.tsv"
raw_data = pd.read_csv(test_path, sep='\t', index_col=0)
hierarchy = ['condition', 'replicate']
# The example data has been background corrected, so set everything below 0 to a trivial positive value of 1
raw_data[raw_data <= 0] = 1
# Remove all genes with low counts so voom isn't confused
raw_data = raw_data[~(raw_data < 5).all(axis=1)]
# Make the Differential Expression Analysis Object
# The reference labels specify how samples will be organized into unique values
dea = DEAnalysis(raw_data,
index_names=hierarchy,
reference_labels=['condition'],
time=None,
counts=True)
# Data can be standarized if desired
# norm_data = dea.standardize()
# Fit the contrasts and save the object
# cont = dea.possible_contrasts()
# cont[0] = 'CRE-BRaf'
dea.fit_contrasts()
dep = DEPlot(dea)
sys.exit()
# Volcano Plot
x = dea.results[0].top_table(p=0.05)
# sns.clustermap(x.iloc[:, :10])
import pandas as pd
from pydiffexp import DEAnalysis
# Load the data
test_path = "/Users/jfinkle/Documents/Northwestern/MoDyLS/Code/Python/sprouty/data/raw_data/all_data_formatted.csv"
raw_data = pd.read_csv(test_path, index_col=0)
hierarchy = ['condition', 'well', 'time', 'replicate']
# The example data has been background corrected, so set everything below 0 to a trivial positive value of 1
raw_data[raw_data <= 0] = 1
# Make the Differential Expression Analysis Object
# The reference labels specify how samples will be organized into unique values
dea = DEAnalysis(raw_data,
index_names=hierarchy,
reference_labels=['condition', 'time'])
# Data can be standarized if desired
# norm_data = dea.standardize()
# Fit the contrasts and save the object
dea.fit_contrasts()
pd.set_option('display.width', 2000)
# print(dea.data)
# print(dea.results['KO-WT'].continuous.loc['NEDD4'])
# print(dea.results['KO-WT'].continuous)
# print(dea.print_experiment_summary())
for key, value in dea.results.items():