def main(opts, mutation_df=None, frameshift_df=None):
# get output file
myoutput_path = opts['output']
opts['output'] = ''
# perform randomization-based test
result_df = rt.main(opts, mutation_df)
# clean up p-values for combined p-value calculation
if opts['kind'] == 'tsg':
p_val_col = 'inactivating p-value'
q_val_col = 'inactivating BH q-value'
elif opts['kind'] == 'effect':
p_val_col = 'entropy-on-effect p-value'
q_val_col = 'entropy-on-effect BH q-value'
elif opts['kind'] == 'oncogene':
p_val_col = 'entropy p-value'
q_val_col = 'entropy BH q-value'
elif opts['kind'] == 'protein':
p_val_col = 'normalized graph-smoothed position entropy p-value'
q_val_col = 'normalized graph-smoothed position entropy BH q-value'
result_df[p_val_col] = result_df[p_val_col].fillna(1)
result_df[q_val_col] = result_df[q_val_col].fillna(1)
if opts['kind'] == 'tsg':
# drop genes that never occur
if opts['kind'] == 'tsg' or opts['kind'] == 'effect':
no_ssvs = (result_df['Total SNV Mutations'] == 0)
result_df = result_df[~no_ssvs]
result_df = result_df.sort_values(by=p_val_col)
elif opts['kind'] == 'oncogene':
# get FDR
result_df = result_df[result_df['Total Mutations'] > 0]
result_df['entropy BH q-value'] = mypval.bh_fdr(
result_df['entropy p-value'])
# combine p-values
result_df['tmp entropy p-value'] = result_df['entropy p-value']
result_df['tmp vest p-value'] = result_df['vest p-value']
result_df.loc[result_df['entropy p-value'] == 0,
'tmp entropy p-value'] = 1. / opts['num_iterations']
result_df.loc[result_df['vest p-value'] == 0,
'tmp vest p-value'] = 1. / opts['num_iterations']
result_df['combined p-value'] = result_df[[
'tmp entropy p-value', 'tmp vest p-value'
]].apply(mypval.fishers_method, axis=1)
result_df['combined BH q-value'] = mypval.bh_fdr(
result_df['combined p-value'])
del result_df['tmp vest p-value']
del result_df['tmp entropy p-value']
if myoutput_path:
# write output if specified
result_df.to_csv(myoutput_path, sep='\t', index=False)
result_df = result_df.set_index('gene', drop=False)
return result_df
def main(opts,
mutation_df=None,
frameshift_df=None):
# get output file
myoutput_path = opts['output']
opts['output'] = ''
# perform randomization-based test
result_df = rt.main(opts, mutation_df)
# clean up p-values for combined p-value calculation
if opts['kind'] == 'tsg':
p_val_col = 'inactivating p-value'
q_val_col = 'inactivating BH q-value'
elif opts['kind'] == 'effect':
p_val_col = 'entropy-on-effect p-value'
q_val_col = 'entropy-on-effect BH q-value'
elif opts['kind'] == 'oncogene':
p_val_col = 'entropy p-value'
q_val_col = 'entropy BH q-value'
elif opts['kind'] == 'protein':
p_val_col = 'normalized graph-smoothed position entropy p-value'
q_val_col = 'normalized graph-smoothed position entropy BH q-value'
elif opts['kind'] == 'hotmaps1d':
p_val_col = 'p-value'
q_val_col = 'q-value'
result_df[p_val_col] = result_df[p_val_col].fillna(1)
result_df[q_val_col] = result_df[q_val_col].fillna(1)
if opts['kind'] == 'tsg':
# drop genes that never occur
if opts['kind'] == 'tsg' or opts['kind'] == 'effect':
no_ssvs = (result_df['Total SNV Mutations']==0)
result_df = result_df[~no_ssvs]
result_df = result_df.sort_values(by=p_val_col)
elif opts['kind'] == 'oncogene':
# get FDR
result_df = result_df[result_df['Total Mutations']>0]
result_df['entropy BH q-value'] = mypval.bh_fdr(result_df['entropy p-value'])
# combine p-values
result_df['tmp entropy p-value'] = result_df['entropy p-value']
result_df['tmp vest p-value'] = result_df['vest p-value']
result_df.loc[result_df['entropy p-value']==0, 'tmp entropy p-value'] = 1. / opts['num_iterations']
result_df.loc[result_df['vest p-value']==0, 'tmp vest p-value'] = 1. / opts['num_iterations']
result_df['combined p-value'] = result_df[['tmp entropy p-value', 'tmp vest p-value']].apply(mypval.fishers_method, axis=1)
result_df['combined BH q-value'] = mypval.bh_fdr(result_df['combined p-value'])
del result_df['tmp vest p-value']
del result_df['tmp entropy p-value']
if myoutput_path:
# write output if specified
result_df.to_csv(myoutput_path, sep='\t', index=False)
result_df = result_df.set_index('gene', drop=False)
return result_df
def handle_oncogene_results(permutation_result, non_tested_genes,
num_permutations):
"""Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
mycols = [
'gene', 'num recurrent', 'position entropy', 'mean vest score',
'entropy p-value', 'vest p-value', 'Total Mutations',
'Unmapped to Ref Tx'
]
permutation_df = pd.DataFrame(permutation_result, columns=mycols)
# get benjamani hochberg adjusted p-values
permutation_df['entropy BH q-value'] = mypval.bh_fdr(
permutation_df['entropy p-value'])
permutation_df['vest BH q-value'] = mypval.bh_fdr(
permutation_df['vest p-value'])
# combine p-values
permutation_df['tmp entropy p-value'] = permutation_df['entropy p-value']
permutation_df['tmp vest p-value'] = permutation_df['vest p-value']
permutation_df.loc[permutation_df['entropy p-value'] == 0,
'tmp entropy p-value'] = 1. / num_permutations
permutation_df.loc[permutation_df['vest p-value'] == 0,
'tmp vest p-value'] = 1. / num_permutations
permutation_df['combined p-value'] = permutation_df[[
'entropy p-value', 'vest p-value'
]].apply(mypval.fishers_method, axis=1)
permutation_df['combined BH q-value'] = mypval.bh_fdr(
permutation_df['combined p-value'])
del permutation_df['tmp vest p-value']
del permutation_df['tmp entropy p-value']
# order output
permutation_df = permutation_df.set_index(
'gene', drop=False) # make sure genes are indices
permutation_df['num recurrent'] = permutation_df['num recurrent'].fillna(
-1).astype(int) # fix dtype isssue
col_order = [
'gene', 'Total Mutations', 'Unmapped to Ref Tx', 'num recurrent',
'position entropy', 'mean vest score', 'entropy p-value',
'vest p-value', 'combined p-value', 'entropy BH q-value',
'vest BH q-value', 'combined BH q-value'
]
permutation_df = permutation_df.sort_values(by=['combined p-value'])
return permutation_df[col_order]
def handle_hotmaps_results(permutation_result):
"""Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
if len(permutation_result[0]) == 6:
mycols = ['gene', 'window length', 'codon position', 'mutation count',
'windowed sum', 'p-value']
else:
mycols = ['gene', 'window length', 'codon position', 'index', 'mutation count',
'windowed sum', 'p-value']
permutation_df = pd.DataFrame(permutation_result, columns=mycols)
# get benjamani hochberg adjusted p-values
permutation_df['q-value'] = 1
for w in permutation_df['window length'].unique():
is_window = permutation_df['window length'] == w
permutation_df.loc[is_window, 'q-value'] = mypval.bh_fdr(permutation_df.loc[is_window, 'p-value'])
#permutation_df['q-value'] = mypval.bh_fdr(permutation_df['p-value'])
# order output
#permutation_df = permutation_df.set_index('gene', drop=False) # make sure genes are indices
col_order = mycols + ['q-value']
permutation_df = permutation_df.sort_values(by=['window length', 'p-value'])
return permutation_df[col_order]
def handle_effect_results(permutation_result):
"""Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
mycols = ['gene', 'num recurrent', 'num inactivating', 'entropy-on-effect',
'entropy-on-effect p-value',
'Total Mutations', 'Unmapped to Ref Tx']
permutation_df = pd.DataFrame(sorted(permutation_result, key=lambda x: x[4] if x[4] is not None else 1.1),
columns=mycols)
# get benjamani hochberg adjusted p-values
permutation_df['entropy-on-effect BH q-value'] = mypval.bh_fdr(permutation_df['entropy-on-effect p-value'])
# order output
permutation_df = permutation_df.set_index('gene', drop=False) # make sure genes are indices
permutation_df['num recurrent'] = permutation_df['num recurrent'].fillna(-1).astype(int) # fix dtype isssue
col_order = ['gene', 'Total Mutations', 'Unmapped to Ref Tx',
'num recurrent', 'num inactivating', 'entropy-on-effect',
'entropy-on-effect p-value', 'entropy-on-effect BH q-value']
return permutation_df[col_order]
def handle_protein_results(permutation_result):
"""Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
mycols = ['gene', 'num recurrent', 'normalized graph-smoothed position entropy',
'normalized graph-smoothed position entropy p-value',
'Total Mutations', 'Unmapped to Ref Tx']
permutation_df = pd.DataFrame(permutation_result, columns=mycols)
# get benjamani hochberg adjusted p-values
permutation_df['normalized graph-smoothed position entropy BH q-value'] = mypval.bh_fdr(permutation_df['normalized graph-smoothed position entropy p-value'])
# order output
permutation_df = permutation_df.set_index('gene', drop=False) # make sure genes are indices
col_order = ['gene', 'Total Mutations', 'Unmapped to Ref Tx',
'num recurrent',
'normalized graph-smoothed position entropy',
'normalized graph-smoothed position entropy p-value',
'normalized graph-smoothed position entropy BH q-value']
permutation_df = permutation_df.sort_values(by=['normalized graph-smoothed position entropy p-value'])
return permutation_df[col_order]
def handle_hotmaps_results(permutation_result):
"""Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
if len(permutation_result[0]) == 5:
mycols = [
'gene', 'codon position', 'mutation count', 'windowed sum',
'p-value'
]
else:
mycols = [
'gene', 'codon position', 'index', 'mutation count',
'windowed sum', 'p-value'
]
permutation_df = pd.DataFrame(permutation_result, columns=mycols)
# get benjamani hochberg adjusted p-values
permutation_df['q-value'] = mypval.bh_fdr(permutation_df['p-value'])
# order output
#permutation_df = permutation_df.set_index('gene', drop=False) # make sure genes are indices
col_order = mycols + ['q-value']
permutation_df = permutation_df.sort_values(by=['p-value'])
return permutation_df[col_order]
def handle_oncogene_results(permutation_result, num_permutations):
"""Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
mycols = ['gene', 'num recurrent', 'position entropy',
'mean vest score', 'entropy p-value',
'vest p-value', 'Total Mutations', 'Unmapped to Ref Tx']
permutation_df = pd.DataFrame(permutation_result, columns=mycols)
# get benjamani hochberg adjusted p-values
permutation_df['entropy BH q-value'] = mypval.bh_fdr(permutation_df['entropy p-value'])
permutation_df['vest BH q-value'] = mypval.bh_fdr(permutation_df['vest p-value'])
# combine p-values
permutation_df['tmp entropy p-value'] = permutation_df['entropy p-value']
permutation_df['tmp vest p-value'] = permutation_df['vest p-value']
permutation_df.loc[permutation_df['entropy p-value']==0, 'tmp entropy p-value'] = 1. / num_permutations
permutation_df.loc[permutation_df['vest p-value']==0, 'tmp vest p-value'] = 1. / num_permutations
permutation_df['combined p-value'] = permutation_df[['entropy p-value', 'vest p-value']].apply(mypval.fishers_method, axis=1)
permutation_df['combined BH q-value'] = mypval.bh_fdr(permutation_df['combined p-value'])
del permutation_df['tmp vest p-value']
del permutation_df['tmp entropy p-value']
# order output
permutation_df = permutation_df.set_index('gene', drop=False) # make sure genes are indices
permutation_df['num recurrent'] = permutation_df['num recurrent'].fillna(-1).astype(int) # fix dtype isssue
col_order = ['gene', 'Total Mutations', 'Unmapped to Ref Tx',
'num recurrent', 'position entropy',
'mean vest score', 'entropy p-value',
'vest p-value', 'combined p-value', 'entropy BH q-value',
'vest BH q-value', 'combined BH q-value']
permutation_df = permutation_df.sort_values(by=['combined p-value'])
return permutation_df[col_order]
def handle_tsg_results(permutation_result):
"""Handles result from TSG results.
Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
permutation_df = pd.DataFrame(sorted(permutation_result,
key=lambda x: x[2]
if x[2] is not None else 1.1),
columns=[
'gene', 'inactivating count',
'inactivating p-value',
'Total SNV Mutations',
'SNVs Unmapped to Ref Tx'
])
permutation_df['inactivating p-value'] = permutation_df[
'inactivating p-value'].astype('float')
tmp_df = permutation_df[permutation_df['inactivating p-value'].notnull()]
# get benjamani hochberg adjusted p-values
permutation_df['inactivating BH q-value'] = np.nan
permutation_df.loc[tmp_df.index,
'inactivating BH q-value'] = mypval.bh_fdr(
tmp_df['inactivating p-value'])
# sort output by p-value. due to no option to specify NaN order in
# sort, the df needs to sorted descendingly and then flipped
permutation_df = permutation_df.sort_values(by='inactivating p-value',
ascending=False)
permutation_df = permutation_df.reindex(index=permutation_df.index[::-1])
# order result
permutation_df = permutation_df.set_index('gene', drop=False)
col_order = [
'gene',
'Total SNV Mutations',
'SNVs Unmapped to Ref Tx',
#'Total Frameshift Mutations', 'Frameshifts Unmapped to Ref Tx',
'inactivating count',
'inactivating p-value',
'inactivating BH q-value'
]
return permutation_df[col_order]
def handle_protein_results(permutation_result):
"""Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
mycols = [
'gene', 'num recurrent', 'normalized graph-smoothed position entropy',
'normalized graph-smoothed position entropy p-value',
'Total Mutations', 'Unmapped to Ref Tx'
]
permutation_df = pd.DataFrame(permutation_result, columns=mycols)
# get benjamani hochberg adjusted p-values
permutation_df[
'normalized graph-smoothed position entropy BH q-value'] = mypval.bh_fdr(
permutation_df[
'normalized graph-smoothed position entropy p-value'])
# order output
permutation_df = permutation_df.set_index(
'gene', drop=False) # make sure genes are indices
col_order = [
'gene', 'Total Mutations', 'Unmapped to Ref Tx', 'num recurrent',
'normalized graph-smoothed position entropy',
'normalized graph-smoothed position entropy p-value',
'normalized graph-smoothed position entropy BH q-value'
]
permutation_df = permutation_df.sort_values(
by=['normalized graph-smoothed position entropy p-value'])
return permutation_df[col_order]
def handle_tsg_results(permutation_result):
"""Handles result from TSG results.
Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
permutation_df = pd.DataFrame(sorted(permutation_result, key=lambda x: x[2] if x[2] is not None else 1.1),
columns=['gene', 'inactivating count', 'inactivating p-value',
'Total SNV Mutations', 'SNVs Unmapped to Ref Tx'])
permutation_df['inactivating p-value'] = permutation_df['inactivating p-value'].astype('float')
tmp_df = permutation_df[permutation_df['inactivating p-value'].notnull()]
# get benjamani hochberg adjusted p-values
permutation_df['inactivating BH q-value'] = np.nan
permutation_df.loc[tmp_df.index, 'inactivating BH q-value'] = mypval.bh_fdr(tmp_df['inactivating p-value'])
# sort output by p-value. due to no option to specify NaN order in
# sort, the df needs to sorted descendingly and then flipped
permutation_df = permutation_df.sort_values(by='inactivating p-value', ascending=False)
permutation_df = permutation_df.reindex(index=permutation_df.index[::-1])
# order result
permutation_df = permutation_df.set_index('gene', drop=False)
col_order = ['gene', 'Total SNV Mutations', 'SNVs Unmapped to Ref Tx',
#'Total Frameshift Mutations', 'Frameshifts Unmapped to Ref Tx',
'inactivating count', 'inactivating p-value',
'inactivating BH q-value']
return permutation_df[col_order]
def handle_effect_results(permutation_result):
"""Takes in output from multiprocess_permutation function and converts to
a better formatted dataframe.
Parameters
----------
permutation_result : list
output from multiprocess_permutation
Returns
-------
permutation_df : pd.DataFrame
formatted output suitable to save
"""
mycols = [
'gene', 'num recurrent', 'num inactivating', 'entropy-on-effect',
'entropy-on-effect p-value', 'Total Mutations', 'Unmapped to Ref Tx'
]
permutation_df = pd.DataFrame(sorted(permutation_result,
key=lambda x: x[4]
if x[4] is not None else 1.1),
columns=mycols)
# get benjamani hochberg adjusted p-values
permutation_df['entropy-on-effect BH q-value'] = mypval.bh_fdr(
permutation_df['entropy-on-effect p-value'])
# order output
permutation_df = permutation_df.set_index(
'gene', drop=False) # make sure genes are indices
permutation_df['num recurrent'] = permutation_df['num recurrent'].fillna(
-1).astype(int) # fix dtype isssue
col_order = [
'gene', 'Total Mutations', 'Unmapped to Ref Tx', 'num recurrent',
'num inactivating', 'entropy-on-effect', 'entropy-on-effect p-value',
'entropy-on-effect BH q-value'
]
return permutation_df[col_order]
