SBML write returns error when values are integers

[paxelito]

Problem description

Hi, while saving a model by using cobra.io.write_sbml_model the system returns the error reported below.

cobra.io.write_sbml_model(some_model, "some_model.sbml")
# Every reaction has the following annotation
{'confidence_level': 2, 'genome_annotation': 'NCBI'} # NOT WORKING
{'confidence_level': [2], 'genome_annotation': 'NCBI'} # IT WORKS

Actual Output

---

TypeError Traceback (most recent call last)
in ()
----> 1 cobra.io.write_sbml_model(myWT5260_NCBI_v1, "Streptomyces-rimosus-WT5260/models/GBM_WT5260_ver1.0.sbml")
2 myWT5260_NCBI_v2 = deepcopy(myWT5260_NCBI_v1)
3 myWT5260_NCBI_v2.model_summary()

~/Documents/git-projects/genome-based-models/env_gbm/lib/python3.6/site-packages/cobra/io/sbml3.py in write_sbml_model(cobra_model, filename, use_fbc_package, **kwargs)
715 return
716 # create xml
--> 717 xml = model_to_xml(cobra_model, **kwargs)
718 write_args = {"encoding": "UTF-8", "xml_declaration": True}
719 if _with_lxml:

~/Documents/git-projects/genome-based-models/env_gbm/lib/python3.6/site-packages/cobra/io/sbml3.py in model_to_xml(cobra_model, units)
512 reversible=str(reaction.lower_bound < 0).lower())
513 set_attrib(sbml_reaction, "name", reaction.name)
--> 514 annotate_sbml_from_cobra(sbml_reaction, reaction)
515 # add in bounds
516 set_attrib(sbml_reaction, "fbc:upperFluxBound",

~/Documents/git-projects/genome-based-models/env_gbm/lib/python3.6/site-packages/cobra/io/sbml3.py in annotate_sbml_from_cobra(sbml_element, cobra_element)
248 if isinstance(identifiers, string_types):
249 identifiers = (identifiers,)
--> 250 for identifier in identifiers:
251 li = SubElement(bag, ns("rdf:li"))
252 set_attrib(li, "rdf:resource", "http://identifiers.org/%s/%s" %

TypeError: 'int' object is not iterable

Expected Output

Save goes successfully

Dependency Information

Please paste the output of python -c "from cobra import show_versions; show_versions()" in the details block below.

System Information

OS Darwin
OS-release 17.5.0
Python 3.6.2

Package Versions

cobra 0.13.0
depinfo 1.3.0
future 0.16.0
numpy 1.14.3
optlang 1.4.2
pandas 0.23.0
pip 9.0.3
ruamel.yaml 0.14.12
setuptools 39.0.1
six 1.11.0
swiglpk 1.4.4
tabulate 0.8.2

[cdiener]

Hi thanks for the report.

Even though it does not say that anywhere in the docs model annotations currently have to be compliant with http://identifiers.org/. So annotations can currently not be anything different from strings. There is an ongoing discussion on how to implement evidence annotations and it is definitely on the radar.

[Midnighter]

@matthiaskoenig is re-working the SBML parser/writer at the moment. I don't think a simple integer will work because the SBML specification requires annotations to be proper XML elements with namespace URI (cf. pp. 15 of the SBML spec PDF http://sbml.org/Documents/Specifications).

For reference you can also take a look at discussions in:

• SBML -> Confidence Scores  schema#9
• SBML ID to Model ID schema#6
• and In cobra.io neither sbml.py nor sbml3.py seem to import or export notes.  schema#4

[paxelito]

First of all thanks for replying so quickly. @Midnighter do you mean that I should add my annotations (to compounds and reactions) in the form like
<mysim:molecule xmlns:mysim="http://www.mysim.org/ns" mysim:weight="18.02" mysim:atoms="3"/>
?

[Midnighter]

Short answer: Not at the moment.

As @cdiener said and as you can see in your above error trace we expect all annotations (except SBO) to resolve to identifiers.org:

"http://identifiers.org/%s/%s" % (provider, identifier)

where 'provider' and 'identifier' are key and value(s) from the 'annotation' dictionary.

However, the SBML spec says that tools should leave annotations that they don't use/can't read alone so it might be an idea to create a separate annotation attribute that contains pure XML. You could then add your own arbitrary XML there. What do you think @matthiaskoenig?

[matthiaskoenig]

@Midnighter I thought in the same direction. Because how things are currently handled all annotations which are not RDF triplets to ontologies are lost during the round trip read - write SBML. I will finish the libsbml issues during the next few days and will have a look at this issue.

But yes, everything which is not annotations must be valid XML.

[matthiaskoenig]

This will be solved as part of #684.