def predict(cv):
col_names = [
line.split(',')[0] for line in open(
'/home/ubuntu/data/scRNAseq/mice/count.csv', 'r').readlines()
][1:]
model = HAL(warm_start=True, n_cluster_init=50, clf_type='rf')
model.load()
ypossible = model.possible_clusters(cv)
X = np.genfromtxt('/home/ubuntu/data/scRNAseq/mice/count.csv',
delimiter=',')
X = X[1:, 1:].T
ypred = model.predict(X, cv)
col_names_all = list(col_names[:, 1].flatten())
df_median_expression = pd.DataFrame(np.array(
[np.median(X[ypred == yu], axis=0) for yu in ypossible]),
index=list(ypossible),
columns=col_names_all)
df_frequency = pd.DataFrame(
[np.count_nonzero(ypred == yu) / len(ypred) for yu in ypossible],
index=ypossible,
columns=[f])
df_frequency.to_csv('/home/ubuntu/data/scRNAseq/mice/Frequencies.csv')
df_median_expression.to_csv(
'/home/ubuntu/data/scRNAseq/mice/Median_expression.csv')
result[f] = [ypred, df_median_expression, df_frequency]
pickle.dump(result, open('results.pkl', 'wb'))
def predict(cv):
file_name_list = []
for filename in os.listdir(
'/Users/mukherjeer2/Documents/Data/CyTOF/20190209_B6_IdU_Pilot/Live_Single_Cells/'
):
if filename.endswith(".fcs"):
file_name_list.append(filename)
continue
else:
continue
col = np.loadtxt('columns.txt', delimiter='\t', dtype=str)
result = {}
model = HAL(warm_start=True, n_cluster_init=50)
model.load()
ypossible = model.possible_clusters(cv)
for f in file_name_list:
print(f)
data = load(f)
data = data[col[:, 0]]
X = np.arcsinh(data)
ypred = model.predict(X, cv)
#Xtmp = model.preprocess(X)
col_names_all = list(col_names[:, 1].flatten())
df_median_expression = pd.DataFrame(np.array(
[np.median(X[ypred == yu], axis=0) for yu in ypossible]),
index=list(ypossible),
columns=col_names_all)
df_frequency = pd.DataFrame(
[np.count_nonzero(ypred == yu) / len(ypred) for yu in ypossible],
index=ypossible,
columns=[f])
df_frequency.to_csv(
'/Users/mukherjeer2/Documents/Data/CyTOF/20190209_B6_IdU_Pilot/Live_Single_Cells/Frequencies.csv'
)
df_median_expression.to_csv(
'/Users/mukherjeer2/Documents/Data/CyTOF/20190209_B6_IdU_Pilot/Live_Single_Cells/Median_expression.csv'
)
#print(df_median_expression)
#print(df_frequency)
#exit()
result[f] = [ypred, df_median_expression, df_frequency]
pickle.dump(result, open('results.pkl', 'wb'))
def analyze(cv):
results = pickle.load(open('results.pkl', 'rb'))
model = HAL(warm_start=True, n_cluster_init=50)
model.load()
#exit()
ypossible = model.possible_clusters(cv=cv)
data = []
idx = []
#df = pd.DataFrame([], columns=[str(yu) for yu in ypossible])
for k, v in results.items():
_, _, df_freq = v
data.append(df_freq.values.flatten())
idx.append(k[:k.find('Singlets.fcs') -
1][find_second_last(k[:k.find('Singlets.fcs') - 1], '_') +
1:])
df = pd.DataFrame(
np.array(data),
index=idx,
columns=results['c11_20190209_BoneMarrow_1-10_01_BM_1_Singlets.fcs']
[2].index)
ax = sns.clustermap(np.arcsinh(df * 1000).T,
xticklabels=df.index,
cbar_kws={'label': 'arcsinh(frequency*1000)'},
figsize=(15, 15))
plt.setp(ax.ax_heatmap.xaxis.get_majorticklabels(),
rotation=90,
ha='right')
plt.setp(ax.ax_heatmap.yaxis.get_majorticklabels(), rotation=00, ha='left')
plt.show()
pickle.dump(df, open('frequencies.pkl', 'wb'))