GeoGrouper

Groups GEO (Gene Expression Omnibus) samples based on the keywords that the sample descriptions share.

Installation

To install the geogrouper package, follow the following steps. The latter steps (5-8) involve the installation of python_mcl, the MCL clustering algorithm implementation in python that this uses.

1. cd <path/to/your/working/directory>
2. git clone https://github.com/mnpatil17/GeoGrouper
3. cd GeoGrouper
4. pip install -e .
5. git clone https://github.com/koteth/python_mcl # do this inside the outer GeoGrouper directory
6. cd python_mcl
7. python setup.py install
8. cd ..

Usage

Using the geogrouper package is very simple. The primary method is cluster_descriptions_from_file

To cluster from a file:

from geogrouper import cluster_descriptions_from_file
clusters_for_each_series = cluster_descriptions_from_file(path_to_data_file)

To cluster from a file AND print to terminal as you go:

from geogrouper import cluster_descriptions_from_file
clusters_for_each_series = cluster_descriptions_from_file(path_to_data_file, should_print_output=True)

To cluster from a file AND print to terminal only the series that have at least N samples:

from geogrouper import cluster_descriptions_from_file
clusters_for_each_series = cluster_descriptions_from_file(path_to_data_file, should_print_output=True, print_series_sample_size=N)

To cluster a list of sample descriptions with some additional description text (abstract_text):

from geogrouper import cluster_descriptions
clusters, mcl_matrix = cluster_descriptions(sample_titles_list, abstract_text)

Modules

1. geo_id.py: handles reading from a specified datatable
2. keywords.py: has multiple methods for finding keywords for a series (not all are used)
3. geogrouper.py: the primary file, which handles the clustering. cluster_descriptions_from_file is the primary method
4. utils.py: various utility functions

Potential Future Optimizations

The file keywords.py contains the logic to find keywords from GEO data. Currently there are two methods:

1. get_acronyms()
2. get_common_words()

Changing the way the main algorithm finds keywords will change the effectiveness of the algorithm. Therefore, to iterate on this algorithm, changing the way keywords are found is a great way to improve performance.