Example for https://www.biostars.org/p/212519/

identify_encode_controls.py

#!/usr/bin/env python

# Install into new environment and activate it before running this script:
#
#   conda create -n biostars-212519 python=3 requests numpy pandas
#   source activate biostars-212519
#   python identify_encode_controls.py

import pandas
import numpy as np
import requests


def accession_metadata(acc):
    """
    Returns the metadata for ENCODE accession `acc` (e.g., ENCSR000BJN)
    """
    HEADERS = {'accept': 'application/json'}
    URL = (
        'https://www.encodeproject.org/experiments/{0}/?frame=object'
        .format(acc)
    )
    response = requests.get(URL, headers=HEADERS)
    return response.content

# You can get a URL interactively on encodeproject.org by subsetting your
# query, clicking on the "download" button, and extracting the first line of
# that file. This example is all HepG2 ChIP-seq data.
metadata_url = (
    "https://www.encodeproject.org/metadata/type=Experiment&"
    "biosample_term_name=HepG2&assay_title=ChIP-seq&limit=all/metadata.tsv"
)

df = pandas.read_table(metadata_url)

# subset just the first 100 rows for this example
df = df.iloc[:100]


def find_controls(acc):
    """
    The metadata for an accession contains a "possible controls" field. I'm
    taking that to mean there can be multiple controls, so to be safe I'm
    returning a list of them.
    """
    m = pandas.read_json(accession_metadata(acc), typ='series')
    c = m['possible_controls']
    return [i.split('/')[2] for i in c]


# the metadata has multiple rows for each accession. To speed things up
# dramatically, only look for controls for the unique set of accessions, and
# then join them to the dataframe afterwards.
#
ds = []
for acc in df['Experiment accession'].unique():
    print('getting metadata for accession:', acc)
    ds.append(
        {
            'Experiment accession': acc,
            'controls': find_controls(acc)
        }
    )
controls = pandas.DataFrame(ds).set_index('Experiment accession')

# join controls to full metadata
df = df.join(controls, on='Experiment accession')