def dmirt(self, cell_lines=['GM12878', 'K562']):
from histogram_cl import histogram_gpu
fpath = os.path.join(self.root, 'prediction', 'pred_tss_20_cnn.db')
con = sqlite3.connect(fpath)
for cell_line in cell_lines:
print(cell_line)
df_ref = pd.read_sql("SELECT miRNA, chromosome, start, end, strand FROM "
"(SELECT * FROM 'upstream_pred_{}')".format(cell_line), con)
df_ref[['start', 'end']] = df_ref[['start', 'end']].astype(int)
tss = df_ref[['start', 'end']].mean(axis=1).astype(int)
df_ref['start'] = tss - 500
df_ref['end'] = tss + 500
gro_path = os.path.join(self.root, 'database', 'GRO_cap.db')
hgpu = histogram_gpu(XmlHandler.load_param("user_param.xml"))
df = hgpu.run(df_ref[['chromosome', 'start', 'end', 'strand']], gro_path, '_'.join([cell_line, 'hg19']))
df['miRNA'] = df_ref['miRNA']
nidx = df[df['strand'] == '-'].index
hist_col = list(range(50))
contents = df.loc[nidx, hist_col[::-1]]
contents.columns = hist_col
df.loc[nidx, hist_col] = contents
dirname = os.path.join(self.root, 'evaluation')
if not os.path.exists(dirname):
os.mkdir(dirname)
out_con = sqlite3.connect(os.path.join(dirname, 'eval_peaks.db'))
df.to_sql('DMIRT_GRO_{}'.format(cell_line), out_con, index=None, if_exists='replace')
def cage_tag(self, paper='DMIRT', cell_lines=['GM12878', 'K562']):
from histogram_cl import histogram_gpu
for cell_line in cell_lines:
if paper == 'DMIRT':
fpath = os.path.join(self.root, 'prediction', 'pred_tss_20_cnn.db')
con = sqlite3.connect(fpath)
df_ref = pd.read_sql("SELECT miRNA, chromosome, start, end, strand FROM "
"(SELECT * FROM 'upstream_pred_{}')".format(cell_line), con)
df_ref[['start', 'end']] = df_ref[['start', 'end']].astype(int)
tss = df_ref[['start', 'end']].mean(axis=1).astype(int)
mir_label = 'miRNA'
elif paper == 'HUA':
fpath = os.path.join(self.root, 'database', 'Supplementary file4-alternative_TSS.db')
con = sqlite3.connect(fpath)
df_ref = pd.read_sql("SELECT * FROM 'cell_specific' WHERE cell_lines LIKE '%{}%'".format(cell_line), con)
tss = df_ref['tss'].astype(int)
mir_label = '#MIR'
elif paper == 'PRO':
fpath = os.path.join(self.root, 'database', 'PRO.db')
con = sqlite3.connect(fpath)
df_ref = pd.read_sql("SELECT * FROM '{}'".format(cell_line), con)
tss = df_ref[['tss_start', 'tss_stop']].mean(axis=1).astype(int)
mir_label = 'miRNA'
else:
fpath = os.path.join(self.root, 'prediction', 'pred_tss_20_cnn.db')
con = sqlite3.connect(fpath)
df_ref = pd.read_sql("SELECT miRNA, chromosome, start, end, strand FROM "
"(SELECT * FROM 'upstream_pred_{}')".format(cell_line), con)
df_ref[['start', 'end']] = df_ref[['start', 'end']].astype(int)
tss = df_ref[['start', 'end']].mean(axis=1).astype(int)
mir_label = 'miRNA'
df_ref['start'] = tss - 500
df_ref['end'] = tss + 500
tss = df_ref[['start', 'end']].mean(axis=1).astype(int)
df_ref['start'] = tss - 500
df_ref['end'] = tss + 500
hgpu = histogram_gpu(XmlHandler.load_param("user_param.xml"))
fpath = os.path.join(self.root, "database", "hCAGE_ctss.db")
rsc_tname = '{}_hg19_ctss'.format(cell_line)
df = hgpu.run(df_ref[['chromosome', 'start', 'end', 'strand']], fpath, rsc_tname)
if mir_label:
df['miRNA'] = df_ref[mir_label]
nidx = df[df['strand'] == '-'].index
hist_col = list(range(50))
contents = df.loc[nidx, hist_col[::-1]]
contents.columns = hist_col
df.loc[nidx, hist_col] = contents
out_con = sqlite3.connect(os.path.join(self.root, 'evaluation', 'eval_peaks.db'))
df.to_sql('{}_CAGE_{}'.format(paper, cell_line), out_con, index=None, if_exists='replace')
def run(self):
import matplotlib.pyplot as plt
dirname__ = os.path.join(self.root, 'database/Histone ChIP-seq')
filenames = XmlHandler.load_param('histogram_param.xml')
N = len(self.histones)
hgpu = histogram_gpu(fpath='histogram_param.xml')
for tissue, fnames in filenames.items():
df_ref = self.load_ref(tissue)
if df_ref is None:
continue
dfs = {}
for histone, fname in fnames.items():
rsc_path = os.path.join(dirname__, tissue, 'histone.db')
hist_out = self.get_hist(df_ref, rsc_path, fname)
print(hist_out)
def hua(self, cell_lines=['GM12878', 'K562']):
from histogram_cl import histogram_gpu
fpath = os.path.join(self.root, 'database', 'Supplementary file4-alternative_TSS.db')
con = sqlite3.connect(fpath)
for cell_line in cell_lines:
df_ref = pd.read_sql("SELECT * FROM 'cell_specific' WHERE cell_lines LIKE '%{}%'".format(cell_line), con)
df_ref['start'] = df_ref['tss'].astype(int) - 500
df_ref['end'] = df_ref['tss'].astype(int) + 500
gro_path = os.path.join(self.root, 'database', 'GRO_cap.db')
hgpu = histogram_gpu(XmlHandler.load_param("user_param.xml"))
df = hgpu.run(df_ref, gro_path, '_'.join([cell_line, 'hg19']))
dirname = os.path.join(self.root, 'evaluation')
if not os.path.exists(dirname):
os.mkdir(dirname)
out_con = sqlite3.connect(os.path.join(dirname, 'eval_peaks.db'))
df.to_sql('HUA_GRO_{}'.format(cell_line), out_con, if_exists='replace', index=None)
def pro(self, cell_lines=['GM12878', 'K562']):
from histogram_cl import histogram_gpu
fpath = os.path.join(self.root, 'database', 'PRO.db')
con = sqlite3.connect(fpath)
for cell_line in cell_lines:
df_ref = pd.read_sql("SELECT * FROM '{}'".format(cell_line), con)
df_ref['tss'] = df_ref[['tss_start', 'tss_stop']].mean(axis=1).astype(int)
df_ref['start'] = df_ref['tss'].astype(int) - 500
df_ref['end'] = df_ref['tss'].astype(int) + 500
gro_path = os.path.join(self.root, 'database', 'GRO_cap.db')
hgpu = histogram_gpu(XmlHandler.load_param("user_param.xml"))
df = hgpu.run(df_ref, gro_path, '_'.join([cell_line, 'hg19']))
dirname = os.path.join(self.root, 'evaluation')
if not os.path.exists(dirname):
os.mkdir(dirname)
out_con = sqlite3.connect(os.path.join(dirname, 'eval_peaks.db'))
df.to_sql('PRO_GRO_{}'.format(cell_line), out_con, if_exists='replace', index=None)