goenrich 1.1.0

GO enrichment with python -- pandas meets networkx

goenrich

Convenient GO enrichments from python. For use in python projects.

1. Builds the GO-ontology graph

2. Propagates GO-annotations up the graph

3. Performs enrichment test for all categories

4. Performs multiple testing correction

5. Allows for export to pandas for processing and graphviz for visualization

Supported ids: Uniport ACC, Entrez GeneID

Installation

Install package from pypi and download ontology

and needed annotations.

pip install goenrich
mkdir db
# Ontology
wget http://purl.obolibrary.org/obo/go/go-basic.obo -O db/go-basic.obo
# UniprotACC
wget http://geneontology.org/gene-associations/gene_association.goa_ref_human.gz -O db/gene_association.goa_ref_human.gz
# Entrez GeneID
wget ftp://ftp.ncbi.nlm.nih.gov/gene/DATA/gene2go.gz -O db/gene2go.gz

Run GO enrichment

import goenrich

# build the ontology
G = goenrich.obo.graph('db/go-basic.obo')

# use all entrez geneid associations form gene2go as background
# use goenrich.read.goa('db/gene_association.goa_ref_human.gz') for uniprot
background = goenrich.read.gene2go('db/gene2go.gz')
goenrich.enrich.set_background(G, background, 'GeneID', 'GO_ID')

# extract some list of entries as example query
query = set(background['GeneID'].unique()[:20])

# run analysis and obtain results
result = goenrich.enrich.analyze(G, query)

# for additional export to graphviz just specify the gvfile argument
# the show argument keeps the graph reasonably small
result = goenrich.enrich.analyze(G, query, gvfile='example.dot', show='top20')

Generate png image using graphviz

dot -Tpng example.dot > example.png

Parameters

Parameters can all be passed to enrich.analyze as shown below

go_options = {
        'multiple-testing-correction' : 'bonferroni',
        'alpha' : 0.05,
        'node_filter' : lambda x : x.get('significant', False)
}
goenrich.enrich.analyze(G, query, **go_options)

# export results to graphviz
goenrich.enrich.analyze(G, query, gvfile='example.dot', **go_options)

Here is an overview over the available parmeters

enrich.analyze:
  node_filter = lambda node : 'p' in node
  show = 'top20' # works for any 'topNUM'

enrich.calculate_pvalues:
  min_category_size = 3
  max_category_size = 500
  min_hit_size = 2

enrich.multiple_testing_correction:
  alpha = 0.05
  method = ['benjamin-hochberg', 'bonferroni']

export.to_frame:
  node_filter = lambda node: True

export.to_graphviz:
  graph_label = None # if None it is replaced by multiple testing info

Licence

This work is licenced under the MIT licence

Contributions are welcome!