def load_immune_landscape(fpath=IMMUNE_LANDSCAPE):
df = (pd.read_csv(fpath).drop('TCGA Study', axis=1).rename(
columns={
'TCGA Participant Barcode': 'patient_id',
'Eosinophils.0': 'Eosinophils score',
'Neutrophils.0': 'Neutrophils score',
'Eosinophils.1': 'Eosinophils',
'Neutrophils.1': 'Neutrophils'
}).dropna(subset=['Wound Healing']).melt(id_vars='patient_id',
var_name='data_type',
value_name='value'))
df.loc[:,
"unit"] = df["data_type"].map(lambda dt: IMMUNE_LANDSCAPE_UNITS[dt])
# Split into numeric values and text values
df['is_numeric'] = df['value'].map(is_numeric)
df_numeric = df[df['is_numeric']]
df_numeric['value'] = df_numeric['value'].map(float)
df_text = df[~df['is_numeric']]
# Load to database in respective tables
conn = db.engine.connect()
df_numeric = (df_numeric[[
'patient_id', 'data_type', 'unit', 'value'
]].drop_duplicates(subset=['patient_id', 'data_type']).dropna(
subset=['value']))
db.copy_from_df(df_numeric, 'patient_value')
df_text = (df_text[[
'patient_id', 'data_type', 'unit', 'value'
]].drop_duplicates(subset=['patient_id', 'data_type']).dropna(
subset=['value']))
db.copy_from_df(df_text, 'patient_text_value')
conn.close()
def load_gtex_median_tpm(fpath=cfg['GTEX_MEDIAN_TPM'], env=cfg['ENV']):
df = (pd.read_csv(fpath, sep='\t',
skiprows=2).melt(id_vars=['gene_id', 'Description'],
var_name='tissue',
value_name='median_tpm'))
df['ensembl_id'] = df['gene_id'].map(lambda s: s.split('.')[0])
if env == 'dev':
df = df[df['Description'].isin(cfg['TEST_SYMBOLS'])]
conn = db.engine.connect()
conn.execute("INSERT INTO source (source_id) VALUES ('GTEx')")
conn.execute('''
CREATE TABLE tmp_gtex
(ensembl_id varchar, tissue varchar, median_tpm numeric)
''')
selected = df[['ensembl_id', 'tissue', 'median_tpm']]
try:
db.copy_from_df(selected, 'tmp_gtex')
conn.execute('''
INSERT INTO tissue_gene_value
(source_id, tissue_id, gene_id, data_type, unit, value)
SELECT 'GTEx', t.tissue_id, g.gene_id, 'expression',
'median_tpm', tmp.median_tpm
FROM tmp_gtex tmp
INNER JOIN tissue t ON t.gtex_id = tmp.tissue
INNER JOIN gene g ON g.ensembl_id = tmp.ensembl_id
''')
finally:
conn.execute('DROP TABLE tmp_gtex')
conn.close()
def load_tcia_patient(fpath=TCIA_PATIENT):
df = (pd.read_csv(fpath, sep='\t', low_memory=False).drop(
['datasource', 'disease'], axis=1).melt(
id_vars='barcode', var_name='data_type',
value_name='value').rename(columns={'barcode': 'patient_id'}))
df.loc[:, 'data_type'] = df['data_type'].str.replace('clinical_data_', '')
df.loc[:, 'unit'] = 'clinical'
# Split into numeric values and text values
df.loc[:, 'is_numeric'] = df['value'].map(is_numeric)
df_numeric = df[df['is_numeric']]
df_numeric.loc[:, 'value'] = df_numeric['value'].map(float)
df_text = df[~df['is_numeric']]
# Load to database in respective tables
df_numeric = (df_numeric[[
'patient_id', 'data_type', 'unit', 'value'
]].drop_duplicates(subset=['patient_id', 'data_type']).dropna(
subset=['value']))
df_text = (df_text[[
'patient_id', 'data_type', 'unit', 'value'
]].drop_duplicates(subset=['patient_id', 'data_type']).dropna(
subset=['value']))
conn = db.engine.connect()
try:
conn.execute('''
CREATE TABLE tmp_tcia_patient_text
(patient_id varchar, data_type varchar,
unit varchar, value varchar)
''')
db.copy_from_df(df_text, 'tmp_tcia_patient_text')
conn.execute('''
INSERT INTO patient_text_value
(patient_id, data_type, unit, value)
SELECT tmp.patient_id, tmp.data_type, tmp.unit, tmp.value
FROM tmp_tcia_patient_text tmp
ON CONFLICT DO NOTHING
''')
conn.execute('''
CREATE TABLE tmp_tcia_patient_num
(patient_id varchar, data_type varchar,
unit varchar, value numeric)
''')
db.copy_from_df(df_numeric, 'tmp_tcia_patient_num')
conn.execute('''
INSERT INTO patient_value
(patient_id, data_type, unit, value)
SELECT tmp.patient_id, tmp.data_type, tmp.unit, tmp.value
FROM tmp_tcia_patient_num tmp
ON CONFLICT DO NOTHING
''')
finally:
conn.execute('DROP TABLE tmp_tcia_patient_text')
conn.execute('DROP TABLE tmp_tcia_patient_num')
conn.close()
def load_tcga_isoforms(fpath, env=cfg['ENV']):
df = (pd.read_csv(fpath).rename(columns={
'transcript_id': 'isoform_id',
'rsem_normalized': 'value'
}))
df['sample_id'] = df['barcode'].map(lambda s: s[:15])
df['unit'] = 'normalized_counts'
df = df[['sample_id', 'isoform_id', 'unit', 'value']]
df = df.drop_duplicates(subset=['sample_id', 'isoform_id'])
if env == 'dev':
df = df[df['isoform_id'].isin(cfg['TEST_ISOFORMS'])]
db.copy_from_df(df, 'sample_isoform_value')
def load_tcia_pathways(up_fpath=TCIA_GSEA_ENRICHMENT,
down_fpath=TCIA_GSEA_DEPLETION):
for fpath in up_fpath, down_fpath:
df = (pd.read_csv(fpath, sep='\t').drop(['disease'], axis=1).rename(
columns={
'patients': 'patient_id',
'cellType': 'data_type',
'qValue < 1%Enriched': 'qvalue',
'qValue < 1%Depleted': 'qvalue',
'NES > 0': 'value'
}))
df = df[df['qvalue'] < 0.05]
df['unit'] = 'normalized enrichment score'
# Load to database in respective tables
df = (df[['patient_id', 'data_type', 'unit', 'value']].drop_duplicates(
subset=['patient_id', 'data_type']).dropna(subset=['value']))
db.copy_from_df(df, 'patient_value')
def load_tcga_clinical(fpath):
mat = pd.read_csv(fpath, sep='\t')
df = mat.melt(id_vars='Hybridization REF',
var_name='patient_id',
value_name='value')
df = df[df['Hybridization REF'] != 'Composite Element REF']
df['patient_id'] = df['patient_id'].str.upper()
df['unit'] = 'clinical'
df = df.rename(columns={'Hybridization REF': 'data_type'})
df = df.drop_duplicates(subset=['patient_id', 'data_type'])
df = df.dropna(subset=['value'])
df['is_numeric'] = df['value'].map(is_numeric)
df_numeric = df[df['is_numeric']]
df_text = df[~df['is_numeric']]
df_numeric = df_numeric[['patient_id', 'data_type', 'unit', 'value']]
db.copy_from_df(df_numeric, 'patient_value')
df_text = df_text[['patient_id', 'data_type', 'unit', 'value']]
db.copy_from_df(df_text, 'patient_text_value')
def load_hpa_protein(fpath=HPA_NORMAL_TISSUE, env=cfg['ENV']):
cell_types = pd.read_csv(CELL_TYPES)
df = pd.read_csv(fpath, sep='\t')
# if env == 'test':
# df = df[df['Gene name'].isin(TEST_GENES)]
formatted = (
df[df['Reliability'] != 'Uncertain']
.merge(
cell_types,
left_on=['Tissue', 'Cell type'],
right_on=['hpa_tissue_id', 'cell_type'])
[['Gene', 'cell_type_id', 'Level']]
.drop_duplicates(subset=['cell_type_id', 'Gene']))
conn = db.engine.connect()
conn.execute("INSERT INTO source (source_id) VALUES ('HPA')")
# Load to temp table
conn.execute('DROP TABLE IF EXISTS tmp_hpa_prot')
conn.execute('''
CREATE TABLE tmp_hpa_prot
(Gene varchar, cell_type_id varchar, Level varchar)
''')
db.copy_from_df(formatted, 'tmp_hpa_prot')
# Insert from temp, joining to existing keys (cell type, gene)
conn.execute('''
INSERT INTO cell_type_gene_text_value
(source_id, cell_type_id, gene_id, data_type, unit, value)
SELECT 'HPA', tmp.cell_type_id, g.gene_id, 'protein',
'detection level', tmp.Level
FROM tmp_hpa_prot tmp
INNER JOIN gene g ON g.ensembl_id = tmp.Gene
''')
conn.close()
def load_hpa_expression(fpath=HPA_RNA_TISSUE, env=cfg['ENV']):
tissues = pd.read_csv(TISSUES)
df = pd.read_csv(fpath, sep='\t')
# if env == 'test':
# df = df[df['Gene name'].isin(TEST_GENES)]
formatted = (
df
.merge(
tissues,
left_on='Sample', right_on='hpa_id')
[['Gene', 'tissue_id', 'Value']]
.drop_duplicates(subset=['tissue_id', 'Gene']))
conn = db.engine.connect()
# Load to temp table
conn.execute('DROP TABLE IF EXISTS tmp_hpa_expr')
conn.execute('''
CREATE TABLE tmp_hpa_expr
(Gene varchar, tissue_id varchar, Value numeric)
''')
db.copy_from_df(formatted, 'tmp_hpa_expr')
# Insert from temp, joining to existing keys (cell type, gene)
conn.execute('''
INSERT INTO tissue_gene_value
(source_id, tissue_id, gene_id, data_type, unit, value)
SELECT 'HPA', tmp.tissue_id, g.gene_id, 'expression',
'TPM', tmp.Value
FROM tmp_hpa_expr tmp
INNER JOIN gene g ON g.ensembl_id = tmp.Gene
''')
conn.execute('DROP TABLE IF EXISTS tmp_hpa_expr')
conn.close()
def load_tcga_sample_meta(cohort_fpath=TCGA_COHORT_META,
patient_fpath=TCGA_PATIENT_META,
sample_fpath=TCGA_SAMPLE_META):
conn = db.engine.connect()
conn.execute("INSERT INTO source (source_id) VALUES ('TCGA')")
cohort = (pd.read_csv(cohort_fpath).assign(source_id='TCGA')[[
'cohort_id', 'source_id', 'cohort_name'
]].drop_duplicates())
db.copy_from_df(cohort, 'cohort')
patient = (pd.read_csv(patient_fpath)[[
'patient_id', 'cohort_id'
]].drop_duplicates(subset=['patient_id']))
db.copy_from_df(patient, 'patient')
sample = (pd.read_csv(sample_fpath)[[
'sample_id', 'patient_id', 'sample_code', 'sample_type'
]].drop_duplicates())
db.copy_from_df(sample, 'sample')
conn.close()
def load_tcga_mutation(fpath=TCGA_MUTATIONS, env=cfg['ENV']):
maf = pd.read_csv(fpath, sep='\t')
#if 'Protein_Change' in maf.columns:
df = maf[[
'Gene',
'Tumor_Sample_Barcode',
'Variant_Classification',
'Variant_Type',
'HGVSp_Short',
]]
# elif 'AAChange' in maf.columns:
# df = maf[
# [
# 'Entrez_Gene_Id',
# 'Tumor_Sample_Barcode',
# 'Variant_Classification',
# 'Variant_Type',
# 'AAChange',
# ]
# ]
# elif 'amino_acid_change' in maf.columns:
# df = maf[
# [
# 'Entrez_Gene_Id',
# 'Tumor_Sample_Barcode',
# 'Variant_Classification',
# 'Variant_Type',
# 'amino_acid_change',
# ]
# ]
# else:
# raise Exception(
# 'MAF file should have either "Protein_Change", "AAChange", '
# 'or "amino_acid_change" column'
# )
df.loc[:, 'sample_id'] = df['Tumor_Sample_Barcode'].map(lambda s: s[:15])
df = df.drop('Tumor_Sample_Barcode', axis=1)
df = df.rename(
columns={
'Gene': 'ensembl_id',
'Variant_Type': 'variant type',
'Variant_Classification': 'variant classification',
'AAChange': 'AA change'
})
# df = df[df['gene_id'] > 0]
# if env == 'dev':
# df = df[df['ensembl_id'].isin(cfg['TEST_ENSEMBL_GENES'])]
df = df.melt(id_vars=['sample_id', 'ensembl_id'],
var_name='data_type',
value_name='value')
df.loc[:, 'unit'] = 'mutation'
df = df.drop_duplicates(subset=['ensembl_id', 'sample_id', 'data_type'])
df = df.dropna(subset=['value'])
df = df[['sample_id', 'ensembl_id', 'data_type', 'unit', 'value']]
conn = db.engine.connect()
try:
conn.execute('''
CREATE TABLE tmp_tcga_maf
(sample_id varchar, ensembl_id varchar, data_type varchar,
unit varchar, value varchar)
''')
db.copy_from_df(df, 'tmp_tcga_maf')
conn.execute('''
INSERT INTO sample_gene_text_value
(sample_id, gene_id, data_type, unit, value)
SELECT tmp.sample_id, g.gene_id, tmp.data_type,
tmp.unit, tmp.value
FROM tmp_tcga_maf tmp
INNER JOIN gene g ON g.ensembl_id = tmp.ensembl_id
INNER JOIN sample s ON s.sample_id = tmp.sample_id
''')
finally:
conn.execute('DROP TABLE tmp_tcga_maf')
conn.close()
def load_gtex_isoform(fpath=cfg['GTEX_MEDIAN_ISOFORM']):
df = (pd.read_csv(fpath).rename(columns={'median_tpm': 'value'}))
df.loc[:, 'unit']: 'TPM'
df = df[['source_id', 'tissue_id', 'isoform_id', 'unit', 'value']]
db.copy_from_df(df, 'tissue_isoform_value')