Expand biosimulators tests that run in docker

[eagmon]

Currently test_biosimulators runs tests on the following biosimulators: biosimulators_tellurium, biosimulators_cobrapy, biosimulators_cbmpy, biosimulators_bionetgen, biosimulators_gillespy2, biosimulators_libsbmlsim, biosimulators_rbapy, biosimulators_xpp.

@jonrkarr -- are there any other simulators we should add to this? Yesterday we mentioned MCell -- is this refactored with exec_sed_task and preprocess_sed_task?

[jonrkarr]

Below are combination possible variations on DFBA. This covers multiple modeling styles, model languages, simulation algorithms, and simulation tools. Within most simulation tools, there are multiple choices for ODE methods. This likely won't be relevant for a relatively simple ODE submodel. But, this capability will be there. This would be relevant for very stiff or otherwise unique ODE systems.

• Constraint-based (steady-state)
  • FBA/SBML
  • biosimulators_cobrapy
  • biosimulators_cbmpy
  • RBA/RBA's own language
  • biosimulators_rbapy
  • Kinetic (uniform time course)
  • ODE
    • SBML
    • biosimulators_amici
    • biosimulators_copasi
    • biosimulators_gillespy2
    • biosimulators_libsbmlsim
    • biosimulators_pysces
    • biosimulators_tellurium
    • biosimulators_vcell -- requires more work from the VCell team
    • Other model languages: models could be converted from SBML
    • biosimulators_bionetgen
    • biosimulators_opencor -- will require a separate Python environment
    • biosimulators_xpp

I suggest starting with FBA/SBML and ODE/SBML and then using other model languages in place of SBML and RBA in place of FBA. For other languages, the SBML version of the FBA model could be converted to BioNetGen, CellML, and XPP. For RBA, we'll need to use a different model which is a superset of the FBA model.

Another variation is to use SSA and similar methods in place of ODE integration (SSA-FBA).

For spatial simulations,

• Smoldyn: Steve has been very responsive, but Smoldyn has several limitations which probably won't be addressed soon. Soon I will try again to push Steve to publish models -- once there's a place to put them.
• VCell: probably the closest to being ready. The VCell group has a Docker image which passes nearly all of our tests. There's no Python API; its implemented in Java. But we could create a Python API from their command-line interface. They may also create a Python API, but this could take a while. We will also have hundreds of VCell models available through BioSimulations soon.
• Morpheus: I've talked to them. They're working on a Python API. Once available, I think they'd be interested to work together with us. We've also discussed contributing their emerging model repository into BioSimulations.
• MCell: I've worked with Jim Faeder's group on BioNetGen and RuleHub. We could talk to them about MCell and a publishing MCell models. I'd like to focus on finishing RuleHub first.

I'm optimistic we'll have better traction with getting the community to contribute once our model repository is available.