def exp_heatmap_json(request):
import pandas as pd
from clustergrammer import Network
columns = [i.name for i in Exp._meta.get_fields()]
#exps = Exp.objects.all().using("expDb").values_list("gene_id", "control_0", "control_1", "control_2", "treated_0", "treated_1", "treated_2")
exps = Exp.objects.all().using("expDb").values()
df = pd.DataFrame(list(exps), columns=columns)
df.index = df.gene_id
df = df.loc[:, df.columns[1:]]
net = Network()
net.load_df(df)
# Z-score normalize the rows
net.normalize(axis='row', norm_type='zscore', keep_orig=True)
# filter for the top 100 columns based on their absolute value sum
net.filter_N_top('col', 100, 'sum')
# cluster using default parameters
net.cluster()
# save visualization JSON to file for use by front end
data = net.export_net_json('viz')
data = json.loads(data)
#print(data)
response = {
'data': data,
}
return JsonResponse(response, content_type='application/json')
def get_clusters(analysis, data_types):
cluster_json = {}
for data_type in data_types:
analysis_data = get_last_analysis_data(analysis, data_type)
data_df, design_df = get_dataframes(analysis_data, PKS[data_type],
SAMPLE_COL)
if data_type == GENOMICS:
inference = WebOmicsInference(data_df, design_df, data_type)
df = inference.standardize_df(inference.data_df)
elif data_type == PROTEOMICS or data_type == METABOLOMICS:
inference = WebOmicsInference(data_df,
design_df,
data_type,
min_value=5000)
df = inference.standardize_df(inference.data_df, log=True)
if not df.empty:
net = Network()
net.load_df(df)
net.cluster()
data_type_label = {
GENOMICS: 'gene',
PROTEOMICS: 'protein',
METABOLOMICS: 'compound'
}
label = data_type_label[data_type]
json_data = net.export_net_json()
cluster_json[label] = json_data
return cluster_json
def build_layout():
df = load_data.load_gsva_compare_cluster('hallmark')
# TODO THIS NEEDS TO BE CLEANED UP!!!!!!
cat_to_true = defaultdict(lambda: [])
for clust in df.index:
if 'PJ030' in clust:
cat_to_true['LGG'].append(clust)
elif 'PJ' in clust:
cat_to_true['GBM'].append(clust)
elif 'LX' in clust:
cat_to_true['LUAD'].append(clust)
elif 'GSE146026' in clust:
cat_to_true['OV'].append(clust)
elif 'GSE72056' in clust:
cat_to_true['SKCM'].append(clust)
elif 'GSE103322' in clust:
cat_to_true['HNSC'].append(clust)
elif 'GSE111672' in clust:
cat_to_true['PAAD'].append(clust)
cats = [{
'title': 'Cancer Type',
'cats': {k: v
for k, v in cat_to_true.items()}
}]
net = Network()
net.load_df(df)
net.add_cats('row', cats)
net.make_clust()
layout = dcc.Tab(label='Cluster Comparison',
children=[
dbc.Container(
fluid=True,
children=[
html.Link(rel='stylesheet',
href='./static/custom.css'),
dash_clustergrammer.DashClustergrammer(
id='cgram-component',
label='',
network_data=net.viz)
])
])
return layout
def get_clustergrammer_json(self, outfile):
# Create network
net = Network()
# Load file
net.load_df(self.expression_dataframe)
# Add categories
try:
net.add_cats('col', self.sample_cats)
except:
pass
try:
# calculate clustering using default parameters
net.cluster()
# save visualization JSON to file for use by front end
net.write_json_to_file('viz', outfile)
except:
os.system('touch {outfile}'.format(**locals()))
def prepare_clustergrammer_data(self,
outfname='clustergrammer_data.json',
G=None):
"""for a distance matrix, output a clustergrammer JSON file
that clustergrammer-js can use
for now it loads the clustergrammer-py module from local dev files
TODO: once changes are pulled into clustergrammer-py, we can use the actual module (pip)
:outfname: filename for the output json
:G: networkx graph (use self.G_sym by default)
"""
G = self.G_sym or self.G
# if Z is None:
# G = self.G_sym or self.G
# Z = self.get_linkage(G)
clustergrammer_py_dev_dir = '../clustergrammer/clustergrammer-py/'
sys.path.insert(0, clustergrammer_py_dev_dir)
from clustergrammer import Network as ClustergrammerNetwork
start = timer()
d = nx.to_numpy_matrix(G)
df = pd.DataFrame(d, index=G.nodes(), columns=G.nodes())
net = ClustergrammerNetwork()
# net.load_file(infname)
# net.load_file(mat)
net.load_df(df)
net.cluster(dist_type='precalculated')
logger.debug("done loading and clustering. took {}".format(
format_timespan(timer() - start)))
logger.debug("writing to {}".format(outfname))
start = timer()
net.write_json_to_file('viz', outfname)
logger.debug("done writing file {}. took {}".format(
outfname, format_timespan(timer() - start)))
new_rows = [(x, 'Cat-1: A', 'Cat-2: B') for x in df.index]
new_cols = [(x, 'Cell Type: Unknown') for x in df.columns]
# new_cols = []
# inst_val = -round(df.shape[1]/2)
# for inst_col in df.columns.tolist():
# inst_val = inst_val + 1
# new_col = (inst_col, 'Cat: C-' + str(inst_val), 'Val: ' + str(inst_val))
# new_cols.append(new_col)
# new_cols = [(x, 'Cat-1: A', 'Cat-2: B') for x in df.columns]
# new_cols = [(x, 'Cat-1: A', 'Cat-2: B', 'Cat-3: C') for x in df.columns]
df.index = new_rows
df.columns = new_cols
net.load_df(df)
net.cluster(dist_type='cos',
views=['N_row_sum', 'N_row_var'],
dendro=True,
sim_mat=False,
filter_sim=0.1,
calc_cat_pval=False,
enrichrgram=True)
# write jsons for front-end visualizations
net.write_json_to_file('viz', 'data/big_data/custom.json', 'no-indent')
# net.write_json_to_file('sim_row', 'json/mult_view_sim_row.json', 'no-indent')
# net.write_json_to_file('sim_col', 'json/mult_view_sim_col.json', 'no-indent')
#make some plots with the combined data
for item in final_LassoLD:
swarms('all', item, featMatAll, directoryA, '.tif' )
plt.close()
#%%Junk from here on
#try clustergrammer
from clustergrammer import Network
net = Network()
net.load_df(cloz_cluster)
# filter for the top 100 columns based on their absolute value sum
net.filter_N_top('col', 16, 'sum')
# cluster using default parameters
net.cluster()
#leg_patch = mpatches.Patch(color = cmap, label=[lut.keys()])
#plt.legend(handles=[leg_patch])
plt.legend ([cmap], lut.keys())
plt.show()
x
swarms (rep1, feature, features_df, directory, file_type)
