Obtain subtypes

get_subtypes.py

import os
import fnmatch
import json
import pandas as pd

def writeout(output_path, df):
    """
        Write dataframe (df) to output path (output_path)
    """
    directory = os.path.dirname(output_path)
    if not os.path.exists(directory):
        os.makedirs(directory)
    df.to_csv(output_path, sep="\t", index=True, header=True)


### Assign a case UUID to each subtype
BASE_DIR = os.path.abspath('/Users/jeffreywong/Sync/bader_jvwong/Guide/datasets/get-data/data/GDC_TCGAOv_Counts')

for file in os.listdir(BASE_DIR):
    if fnmatch.fnmatch(file, 'metadata.cart*.json'):
        metadata_file = os.path.join(BASE_DIR, file)
subtypes_file = os.path.join(BASE_DIR, 'Verhaak_JCI_2013_tableS1.txt')
output_dir = os.path.join(BASE_DIR, 'output')
output_file_subtype = os.path.join(output_dir, 'TCGAOv_subtypes.txt')

df_subtypes = pd.read_table(subtypes_file, index_col=0, header=0)
df_gdc = pd.DataFrame(columns=['case_id'])

subtypes = ['Mesenchymal', 'Immunoreactive']

with open(metadata_file, 'r') as f:
    metadatas = json.load(f)

for idx, metadata in enumerate(metadatas):
    if not "associated_entities" in metadata or not metadata["associated_entities"]:
        continue

    ## debug notes - is this mapping right???
    case_id = metadata["associated_entities"][0]["case_id"]
    entity_submitter_id = metadata["associated_entities"][0]["entity_submitter_id"]
    ## Note 1
    barcode = '-'.join(entity_submitter_id.split('-')[0:3])

    # Note 2
    if barcode in df_subtypes.index and df_subtypes.ix[barcode]['SUBTYPE'] not in subtypes:
        continue
    df_gdc.ix[barcode] = case_id

## Note 3
df_assigned = pd.merge(df_subtypes, df_gdc,
                       how='inner',
                       left_index=True,
                       right_index=True)
df_assigned.index.name = 'barcode'

## Note 4
df_assigned.sort_values('case_id', inplace=True)
writeout(output_file_subtype, df_assigned)