PyENCODE 0.1

Convenience package for accessing ENCODE (Encyclopedia of DNA Elements) project data at UCSC

This is a convenience package for accessing the raw data of the ENCODE (Encyclopedia of DNA Elements) project.

The raw ENCODE files are organized in a fairly straightforward structure under this URL. The files are divided into collections (“composites”), each in its own subdirectory. The subdirectory of each collection keeps the metadata of all files in a text file named files.txt. For example, the genome segmentation data is kept under ROOT_URL/wgEncodeAwgSegmentation. In particular, the segmentation obtained using Combined method on the K562 cells is kept in a compressed BED-file named ROOT_URL/wgEncodeAwgSegmentation/wgEncodeAwgSegmentationCombinedK562.bed.gz.

In principle, downloading and reading the file is rather straightforward. What this package offers in addition is a slightly more streamlined way of listing files, caching them, and reading file metadata. For example, the following code will download the abovementioned file into cache, then open it and index as an interval tree:

>> from pyencode import Encode
>> e = Encode(cache_dir = 'wgEncode')
>> gtree = e.AwgSegmentation.CombinedK562.fetch().read_as_intervaltree()

As another example, here is how to list and pre-download all files in the AwgSegmentation collection into cache:

>> for f in e.AwgSegmentation:
>>    print("%s-%s" % (f['cell'], f['dataType']))
>>    f.fetch()

Installation

The easiest way to install most Python packages is via easy_install or pip:

$ pip install PyENCODE

Usage

The main object, provided by the package is pyencode.Encode. You create an instance, specifying the root of a cache directory:

>> from pyencode import Encode
>> e = Encode(cache_dir = 'wgEncode')

The default value for cache_dir is ~/.pyencode. The resulting object works as a dictionary, with keys being the different file collections within ENCODE:

>> c['AwgSegmentation']

Alternatively, you can use field names instead of dictionary keys, i.e. e['AwgSegmentation'] is the same as e.AwgSegmentation. To iterate over all collections, simply do:

>> for c in e:
>>    print(c.name)

Each element of the Encode object is a EncodeCollection object, which is acts as a collection of EncodeFile elements:

>> for f in e.AwgSegmentation:
>>     print(f.name)

Simiarly, dictionary-style or field name access can be used to retrieve files in a collection: e.AwgSegmentation['CombinedK562'] or e.AwgSegmentation.CombinedK562.

Each EncodeFile is a dictionary of file metadata fields:

>> print(e.AwgSegmentation.CombinedK562['cell'])

In addition, EncodeFile provides a set of convenience fields and methods:

• fetch(force=False) - Download file into cache. Returns the EncodeFile object for convenient chaining of calls. When``force`` is False, file will not be redownloaded if already in cache.

• keys() - Set of all file attributes that can be accessed via [].

• url - Return the URL of the file online.

• local_url - The URL of the cached copy. It is not guaranteed that the file exists, so it is often more practical to do .fetch().local_url.

• local_path - Return the path of the locally cached copy. It is not guaranteed that the file exists.

• open() - Open the file in binary mode for reading. If the file is not in cache, it is not downloaded to cache and opened from the web (so, it is often more practical to do .fetch().open()).

• open_text() - Open the file in text mode for reading. If the file is not in cache it is not downloaded to cache and opened from the web. If the file is a .gz file, it is automatically unpacked (i.e. the returned file instance is an opened GzipFile).

• read_as_intervaltree() - Read a BED file into an intervaltree.bio.GenomeIntervalTree data structure. Simiarly, if the file is not in cache, it is not automatically downloaded.

Note that Encode is not safe for doing multithreading or multiprocessing, unless all the necessary files are already cached.

Copyright & License

Copyright 2014, Konstantin Tretyakov

MIT License.

Submit issues and fixes at https://github.com/konstantint/PyENCODE