def get_TCGA_cnv(self, input_list):
samples = xena.cohort_samples(self.host, self.cohort, None) #### All Sample selection
#### Create ID mapper
pos_df = pd.read_csv('TCGAlib/dataset/position_info_hg38.csv', index_col=0)
#### ID mapping
id_map = self.id_mapper_to_entrez(pos_df, input_list)
#### Get Ensembl ID from EntrezID
id_map['chr'] = id_map['chr'].apply(lambda x:'chr'+x)
id_map['start'].astype(int)
id_map['end'].astype(int)
#### Retreive values from Xena
values = [xena.segmented_data_range(self.host, self.cnv_dataset, samples, id_map['chr'].loc[el], id_map['start'].loc[el], id_map['end'].loc[el]) for el in id_map.index.tolist()]
cnv_arr = []
for i,item in enumerate(input_list):
temp_df = pd.DataFrame(data=values[i]['rows']['value'], columns=[item], index=values[i]['rows']['sampleID'])
if not temp_df.empty:
####There are duplicated samples in Xena, don't know the reason.
temp_df = temp_df[~temp_df.index.duplicated(keep='first')].loc[samples]
cnv_arr.append(temp_df)
if len(input_list)==1:
cnv_df = cnv_arr[0].astype(float)
else:
cnv_df = pd.concat(cnv_arr, axis=1).astype(float)
return cnv_df
def tutorial_main():
## TCGA Hub Data
host = xena.PUBLIC_HUBS['tcgaHub']
excludeTypes = ['probeMap', 'probemap', 'genePredExt']
cohort_list = xena.all_cohorts(host, excludeTypes)
cohort_symbol_list = [x.split('(')[1][:-1] for x in cohort_list]
# print(cohort_symbol_list)
cohort= 'TCGA Bile Duct Cancer (CHOL)'
cohort_symbol = 'CHOL'
# cohort = "GDC TCGA Bile Duct Cancer (CHOL)"
# cohort_symbol = 'CHOL'
samples = xena.cohort_samples(host, cohort, None)
## mode = 'all' , 'selected'
mode = 'selected'
genes = ['ITGB5']
rna_expressions_by_genes = get_TCGA_rna_expressions(host, samples, cohort_symbol,mode,genes)
print(rna_expressions_by_genes[genes[0]])
# Overall Survival data
os_phenotypes = get_TCGA_OS_phenotypes(host, samples, cohort_symbol)
os_result = clean_samples_for_tumour(rna_expressions_by_genes[genes[0]], os_phenotypes)
os_time = pd.to_numeric(os_result['os_time'],errors='coerce')
os_event = pd.to_numeric(os_result['os_event'], errors='coerce')
rna_expression = list(os_result['expression'])
os_groups = get_TCGA_os_groups(os_result['expression'], os_group_type ='quartiles')
print(os_groups)
os_df = pd.DataFrame(
{
'os_time' : os_time,
'os_group': os_groups,
'os_event': os_event,
}
)
os_df = make_df_for_kaplan(os_result, 'quartiles')
# os_df = make_df_for_kaplan(os_result, 'median')
print(os_df.head())
results = multivariate_logrank_test(os_df['os_time'],os_df['os_group'],os_df['os_event'])
print (results)
print(results.p_value)
print(len(os_df))
def kaplan_meier_a_gene_all_TCGA(genes,grouping_mode, result_dir):
## TCGA Hub Data
host = xena.PUBLIC_HUBS['tcgaHub']
excludeTypes = ['probeMap', 'probemap', 'genePredExt']
cohort_list = xena.all_cohorts(host, excludeTypes)
cohort_list.remove('TCGA Pan-Cancer (PANCAN)')
cohort_list.remove('TCGA Formalin Fixed Paraffin-Embedded Pilot Phase II (FPPP)')
cohort_symbol_list = [x.split('(')[1][:-1] for x in cohort_list]
# cohort = 'TCGA Bile Duct Cancer (CHOL)'
# cohort_symbol = 'CHOL'
tcga_os_kaplan_result_dict = dict()
#################
## Temp
#################
# cohort_list = ['TCGA Bile Duct Cancer (CHOL)']
# cohort_symbol_list =['CHOL']
for cohort in cohort_list:
cohort_symbol = cohort.split('(')[1][:-1]
samples = xena.cohort_samples(host, cohort, None)
# print(cohort, cohort_symbol)
## mode = 'all' , 'selected'
mode = 'selected'
rna_expressions_by_genes = get_TCGA_rna_expressions(host, samples, cohort_symbol, mode, genes)
# Overall Survival data
os_phenotypes = get_TCGA_OS_phenotypes(host, samples, cohort_symbol)
# remove NaN samples, and select only tumour samples
os_result = clean_samples_for_tumour(rna_expressions_by_genes[genes[0]], os_phenotypes)
os_df = make_df_for_kaplan(os_result, grouping_mode)
# print("sample:",len(os_df))
results = multivariate_logrank_test(os_df['os_time'], os_df['os_group'], os_df['os_event'])
# print("p-value:",results.p_value)
tcga_os_kaplan_result_dict[cohort] = results.p_value
# for key,value in tcga_os_kaplan_result_dict.items():
# print(key, value)
tcga_os_kaplan_result_df = pd.DataFrame.from_dict(tcga_os_kaplan_result_dict, orient='index')
tcga_os_kaplan_result_df.to_csv("{}/{}_kaplan_meier_all_TCGA.tsv".format(result_dir, genes[0]),
sep='\t')
def kaplan_meier_all_gene_a_TCGA(cohort, grouping_mode, result_dir):
## TCGA Hub Data
host = xena.PUBLIC_HUBS['tcgaHub']
excludeTypes = ['probeMap', 'probemap', 'genePredExt']
# cohort_list = xena.all_cohorts(host, excludeTypes)
cohort_symbol = cohort.split('(')[1][:-1]
rna_seq_dataset = "TCGA.{}.sampleMap/HiSeqV2_PANCAN".format(cohort_symbol)
all_genes = xena.dataset_field(host, rna_seq_dataset) ## all gene sets
# for a_cohort in cohort_list:
# if cohort_symbol in a_cohort:
# cohort = a_cohort
samples = xena.cohort_samples(host, cohort, None)
mode = 'all'
rna_expressions_by_genes = get_TCGA_rna_expressions(host, samples, cohort_symbol, mode)
tcga_os_kaplan_result_dict = dict()
for gene in all_genes:
print("cohort , gene", cohort, gene)
## mode = 'all' , 'selected'
mode = 'selected'
# Overall Survival data
os_phenotypes = get_TCGA_OS_phenotypes(host, samples, cohort_symbol)
# remove NaN samples, and select only tumour samples
os_result = clean_samples_for_tumour(rna_expressions_by_genes[gene], os_phenotypes)
os_df = make_df_for_kaplan(os_result, grouping_mode)
# print("sample:", len(os_df))
results = multivariate_logrank_test(os_df['os_time'], os_df['os_group'], os_df['os_event'])
# print("p-value:", results.p_value)
tcga_os_kaplan_result_dict[gene] = results.p_value
for key, value in tcga_os_kaplan_result_dict.items():
print(key, value)
tcga_os_kaplan_result_df = pd.DataFrame.from_dict(tcga_os_kaplan_result_dict, orient='index')
tcga_os_kaplan_result_df.to_csv("{}/{}_kaplan_meier_all_genes.tsv".format(result_dir, cohort_symbol),sep='\t')
def get_TCGA_expr(self, input_list):
samples = xena.cohort_samples(self.host, [self.cohort], None) #### All Sample selection
#### Create ID mapper
xena_transcript = xena.dataset_field(self.host, self.expr_dataset) #### All Transcripts
xena_transcript_df = pd.DataFrame(data=xena_transcript, columns =['EnsemblID'])
xena_transcript_df['EnsemblID_edit'] = xena_transcript_df['EnsemblID'].apply(lambda x : x.split('.')[0])
xena_transcript_df = xena_transcript_df.set_index('EnsemblID_edit')
#### ID mapping & Get Ensembl ID from EntrezID
id_map = self.id_mapper_to_entrez(xena_transcript_df, input_list)
#### Retreive values from Xena
values = xena.dataset_fetch(self.host, self.expr_dataset, samples, id_map['EnsemblID'].values.tolist()) # list of lists
#xena_df = pd.DataFrame(data=values, index=input_list, columns=samples)
xena_df = pd.DataFrame(data=values, index=input_list, columns=samples).T.astype(float)
return xena_df
def get_TCGA_mut(self, input_list):
samples = xena.cohort_samples(self.host, self.cohort, None) #### All Sample selection
#### Create ID mapper
pos_df = pd.read_csv('TCGAlib/dataset/position_info_hg38.csv', index_col=0)
#### ID mapping
id_map = self.id_mapper_to_entrez(pos_df, input_list)
#### Get Ensembl ID from EntrezID
id_map['chr'] = id_map['chr'].apply(lambda x:'chr'+x)
id_map['start'].astype(int)
id_map['end'].astype(int)
#### Retreive values from Xena
values = [xena.sparse_data_range(self.host, self.mut_dataset, samples, id_map['chr'].loc[el], id_map['start'].loc[el], id_map['end'].loc[el]) for el in id_map.index.tolist()]
result_arr = [pd.DataFrame(data=[values[i]['rows']['amino-acid'], values[i]['rows']['effect']], columns=values[i]['rows']['sampleID'], index=[item+'_AAC', item+'_Effect']).T for i,item in enumerate(id_map.index.tolist())]
#mut_df = pd.concat(result_arr, axis=1)
return result_arr