Handling and writing OBO
Project description
Tools for biological identifiers, names, synonyms, xrefs, hierarchies, relations, and properties through the perspective of OBO.
Example Usage
Note! PyOBO is no-nonsense. This means that there’s no repetitive prefixes in identifiers. It also means all identifiers are strings, no exceptions.
Note! The first time you run these, they have to download and cache all resources. We’re not in the business of redistributing data, so all scripts should be completely reproducible. There’s some AWS tools for hosting/downloading pre-compiled versions in pyobo.aws if you don’t have time for that.
Note! PyOBO can perform grounding in a limited number of cases, but it is not a general solution for named entity recognition (NER) or grounding. It’s suggested to check Gilda for a no-nonsense solution.
Mapping Identifiers and CURIEs
Get mapping of ChEBI identifiers to names.
import pyobo
chebi_id_to_name = pyobo.get_id_name_mapping('chebi')
name = chebi_id_to_name['132964']
assert name == 'fluazifop-P-butyl'Or, you don’t have time for two lines
import pyobo
name = pyobo.get_name('chebi', '132964')
assert name == 'fluazifop-P-butyl'Get reverse mapping of ChEBI names to identifiers
import pyobo
chebi_name_to_id = pyobo.get_name_id_mapping('chebi')
identifier = chebi_name_to_id['fluazifop-P-butyl']
assert identifier == '132964'Maybe you live in CURIE world and just want to normalize something like CHEBI:132964:
import pyobo
name = pyobo.get_name_by_curie('CHEBI:132964')
assert name == 'fluazifop-P-butyl'Remap a CURIE based on pre-defined priority list and Inspector Javert’s Xref Database:
import pyobo
# Map to the best source possible
mapt_ncbigene = pyobo.get_priority_curie('hgnc:6893')
assert mapt_ncbigene == 'ncbigene:4137'
# Sometimes you know you're the best. Own it.
assert 'ncbigene:4137' == pyobo.get_priority_curie('ncbigene:4137')Grounding
Maybe you’ve got names/synonyms you want to try and map back to ChEBI synonyms. Given the brand name Fusilade II of CHEBI:132964, it should be able to look it up and its preferred label.
import pyobo
prefix, identifier, name = pyobo.ground('chebi', 'Fusilade II')
assert prefix == 'chebi'
assert identifier == '132964'
assert name == 'fluazifop-P-butyl'
# When failure happens...
prefix, identifier, name = pyobo.ground('chebi', 'Definitely not a real name')
assert prefix is None
assert identifier is None
assert name is NoneIf you’re not really sure which namespace a name might belong to, you can try a few in a row (prioritize by ones that cover the appropriate entity type to avoid false positives in case of conflicts):
import pyobo
# looking for phenotypes/pathways
prefix, identifier, name = pyobo.ground(['efo', 'go'], 'ERAD')
assert prefix == 'go'
assert identifier == '0030433'
assert name == 'ubiquitin-dependent ERAD pathway'Cross-referencing
Get xrefs from ChEBI to PubChem
import pyobo
chebi_id_to_pubchem_compound_id = pyobo.get_filtered_xrefs('chebi', 'pubchem.compound')
pubchem_compound_id = chebi_id_to_pubchem_compound_id['132964']
assert pubchem_compound_id == '3033674'Get xrefs from Entrez to HGNC, but they’re only available through HGNC so you need to flip them
import pyobo
hgnc_id_to_ncbigene_id = pyobo.get_filtered_xrefs('hgnc', 'ncbigene')
ncbigene_id_to_hgnc_id = {
ncbigene_id: hgnc_id
for hgnc_id, ncbigene_id in hgnc_id_to_ncbigene_id.items()
}
mapt_hgnc = ncbigene_id_to_hgnc_id['4137']
assert mapt_hgnc == '6893'Since this is a common pattern, there’s a keyword argument flip that does this for you:
import pyobo
ncbigene_id_to_hgnc_id = pyobo.get_filtered_xrefs('hgnc', 'ncbigene', flip=True)
mapt_hgnc = ncbigene_id_to_hgnc_id['4137']
assert mapt_hgnc == '6893'Properties and Relations
Get properties, like SMILES. The semantics of these are defined on an OBO-OBO basis.
import pyobo
# I dont make the rules. I wouldn't have chosen this as the key for this property. It could be any string
chebi_smiles_property = 'http://purl.obolibrary.org/obo/chebi/smiles'
chebi_id_to_smiles = pyobo.get_filtered_properties_mapping('chebi', chebi_smiles_property)
smiles = chebi_id_to_smiles['132964']
assert smiles == 'C1(=CC=C(N=C1)OC2=CC=C(C=C2)O[C@@H](C(OCCCC)=O)C)C(F)(F)F'Check if an entity is in the hierarchy:
import networkx as nx
import pyobo
# check that go:0008219 ! cell death is an ancestor of go:0006915 ! apoptotic process
assert 'go:0008219' in pyobo.get_ancestors('go', '0006915')
# check that go:0070246 ! natural killer cell apoptotic process is a
# descendant of go:0006915 ! apoptotic process
apopototic_process_descendants = pyobo.get_descendants('go', '0006915')
assert 'go:0070246' in apopototic_process_descendantsGet the subhierarchy below a given node:
# get the descendant graph of go:0006915 ! apoptotic process
apopototic_process_subhierarchy = pyobo.get_subhierarchy('go', '0006915')
# check that go:0070246 ! natural killer cell apoptotic process is a
# descendant of go:0006915 ! apoptotic process through the subhierarchy
assert 'go:0070246' in apopototic_process_subhierarchyGet a hierarchy with properties pre-loaded in the node data dictionaries:
import pyobo
prop = 'http://purl.obolibrary.org/obo/chebi/smiles'
chebi_hierarchy = pyobo.get_hierarchy('chebi', properties=[prop])
assert 'chebi:132964' in chebi_hierarchy
assert prop in chebi_hierarchy.nodes['chebi:132964']
assert chebi_hierarchy.nodes['chebi:132964'][prop] == 'C1(=CC=C(N=C1)OC2=CC=C(C=C2)O[C@@H](C(OCCCC)=O)C)C(F)(F)F'Installation 
PyOBO can be installed from PyPI with:
$ pip install pyoboIt can be installed in development mode from GitHub with:
$ git clone https://github.com/pyobo/pyobo.git
$ cd pyobo
$ pip install -e .Curation of the Metaregistry
At src/pyobo/registries/metaregistry.json is the curated registry. This is a source of information that contains all sorts of fixes for missing/wrong information in MIRIAM, OLS, and OBO Foundry; entries that don’t appear in any of them; additional synonym information for each namespace/prefix; rules for normalizing xrefs and CURIEs, etc.
Most users will be interested in the "database" subdictionary. Each entry has a key that was chosen first by preference for MIRIAM, then OBO Foundry, then OLS, or assigned based on what felt right/was how they appeared in xrefs in other OBO files. Their corresponding entries can have some combination of these keys:
title
pattern, a regex string for identifiers
url, a url pattern to resolve identifiers. Uses $1 to represent the identifier.
synonyms, a list of alternative prefixes that should point to this
download, a URL to the OBO file in case OBO Foundry doesn’t list it or has a mistake
not_available_as_obo, a boolean telling you exactly what it sounds like
no_own_terms, a boolean telling you if it is completely derived from external sources
wikidata_property, a string pointing to the wikidata property that connects item in WikiData to identifers in this namespace
miriam: a dictionary containing “id” and “prefix” to point to MIRIAM
obofoundry: a dictionary containing “prefix” to point to OBO Foundry
ols, a dictionary containing “ontologyId” to point to OLS
Other entries in the metaregistry:
The "remappings"->"full" entry is a dictionary from strings that might follow xref: in a given OBO file that need to be completely replaced, due to incorrect formatting
The "remappings"->"prefix" entry contains a dictionary of prefixes for xrefs that need to be remapped. Several rules, for example, remove superfluous spaces that occur inside CURIEs or and others address instances of the GOGO issue.
The "obsolete" entry maps prefixes that have been changed.
The "blacklists" entry contains rules for throwing out malformed xrefs based on full string, just prefix, or just suffix.
Troubleshoting
The OBO Foundry seems to be pretty unstable with respect to the URLs to OBO resources. If you get an error like:
pyobo.getters.MissingOboBuild: OBO Foundry is missing a build for: mondoThen you should check the corresponding page on the OBO Foundry (in this case, http://www.obofoundry.org/ontology/mondo.html) and make an update to the url entry for that namespace in the metaregistry.
Development
Update the registries with the following commands. These external resources get updated all the time, so don’t forget about this.
$ python -m pyobo.registries.obofoundry
$ python -m pyobo.registries.ols
$ python -m pyobo.registries.miriamRelease history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
