A collection of python tools. These mainly involve:

1. Data processing
2. Networks (graphs) and related algorithms

helpers for pandas

• split_and_stack, which creates multiple rows from string delimited columns
• merge, which is an in-place replacement for pandas merge (can also be assigned to pandas.DataFrame.merge) which is slightly better (in my opinion, see the docs :-))
• data_uri() helps to generate links to data in ipython notebook (avoid using large data frames). The basics:

    from IPython.display import HTML
    df = pandas.DataFrame(...)
    data_uri(df, HTML)

Helper methods to read (and apply common processing) for some prevalent file formats. Generally methods allow the client to specify additional arguments that are passed directly to the underlying pandas parser. Currently the following are available:

1. gene_info (NCBI).
2. uniprot mapping. Will attempt to build a mapping table between the requested formats, based on a uniprot mapping file
3. BioGRID. Will read tab2 files from biogrid. There's a general biogrid function but biogrid_physical and biogrid_genetic may be more suitable for reading only physical (genetic) interactions.
4. ANAT. Read networks from ANAT.
5. HIPPIE. Read HIPPIE network.
6. GOA (GO association) file format.

Adds some convenience plotting functions:

    corr_plot(df, method='spearman', **kwargs)
    heat_plot(df, cluster=False, colorbar=True, **kwargs)
    na_plot(df, cluster=False)
    auc_plot(scores, truth_values, label_template=None, **kwargs)

A very useful SparseGraph class that behaves as a mixture of a sparse-matrix and a DataFrame. Additionally, there are a few algorithms in this package, some are better implemented than their scipy counterparts.

random graph generation (geometric and preferential attachment) and degree-preserving shuffling (requires a compilation of the accompanying c library)

Convenient cross-platform directory creation with won't blow up in you face

Statistical utilities. Currently contains only partial correlation

Currently contains Jaccard pairwise distance for sparse matrices (much faster than the dense version that is shipped with scipy)

Exposes the KEGG database as a set of pandas.DataFrame\s

Allows for fast and convenient annotation of other pandas.DataFrames using the merge and groupby operations.

The database is initially downloaded into the path defined in kegg/settings.py. For subsequent import the already downloaded tables are read from cache.

Available dataframes:

# raw
pathways
entries
reactions
relations

# processed
nodes
edges

Note that I currently don't package it. I just clone the repository directly into: '<my_virtual_env>/lib/python2.7/site-packages'

If you don't know where it is run

>>> import site
>>> site.getsitepackages()[0]