ABI Meeting Minutes 2008-05-28

Present: Catherine Lloyd, James Lawson, Tommy Yu, Justin Marsh, Mike Cooling, Geoff Nunns, Edmund Crampin, Poul Nielsen, Andrew Miller

Review of last week's action items:

Action item 1.) James to investigate

Agenda for this week:

Geoff Nunns starting internship

• Geoff Nunns is from Johns Hopkins University in Baltimore, USA
• He is starting an internship under Catherine and will be working here until August
• He will primarily be working on coding up new CellML models
  

• Mike has created three tutorials on how to work with CellML files for the Newcastle iGEM team
• The first deals with using COR to create a simple one-component model with some mass-action kinetics math
  
• The second deals with a two-component model and connections
• The third deals with imports and PCEnv
  
• He wondered if it would be a good idea to make these tutorials available via www.cellml.org
• It would be good to let the Newcastle group, who don't know much about CellML, trial the tutorials first
• Geoff could also check out these tutorials, as he is just learning how to work with CellML
• Mike has written the tutorials in Word. The format of the tutorials on the website was discussed
• A wiki was suggested to allow easy editing, or they could be converted into a Plone 'page,' which would also allow editing
  
• Mike isn't keen to spend time translating the Word documents into another format

• PCEnv 0.4 release will be coming up soon, but the documentation needs to be updated
• This shouldn't be too much work, but screenshots need to be replaced and some new features need to be documented
• James and Catherine worked on this for PCEnv 0.3 - Randall wanted to know if they could do it this time as well.
• James is happy to work on this with Justin

• Justin has been working on the bugs marked as blocking the release of PCEnv 0.4
• Andrew has code generation for differential algebraic equation (DAE) systems operational
• Andrew is still looking into different minimisation / root-finding methods to be called from the generated code.
• He was able to generate code for the Saucerman & McCulloch 2003 model - 6 pages of code were generated and 11 DAE's were found by the system
• Poul asked if attendees of last week's seminar on DAEs received any useful feedback
• Randall asked about the validity of using separate solvers for DAEs and ODEs as a numerical approach for solving models that contain both. Apparently separating DAEs from ODEs using separate solvers on them can reduce error handling capabilities.
• Poul noted that the "RADAU5" solver would be worth looking into
• Andrew said this solver was used in MozCellML
• Edmund mentioned that the "NAG" libraries would also be worth checking out. They are closed source but should give a good sense of the current state of the art.
• John Butcher in the University of Auckland Department of Mathematics is a world expert on numerical methods for the solution of ODEs and would definitely be a good person to talk to about our work in this field.
• Randall suggested that if we have the analytical solution to an appropriate problem, we could compare this with our numerical solution.
• GAMS (Guide to Available Mathematical Software) is a resource that may be relevant.
• Randall mentioned that Andrew's current software could be used for MATLAB code generation with not too much modification.

• Tommy solicited comments from various people after last week's meeting
• He has been working on cleaning up the interface to get the prototype ready for preliminary user testing
• The VM should be ready for distribution and hosting by tomorrow.
• He has also built and played around with Andre's 'decompose' tool.
• The Ten Tusscher 2004 model variants do not quite have identical mathematics and variable units. This complicates using imports to create a one main model that imports different parameter sets to describe the variants.
• James mentioned that there are some strange units such as milliamps per picoamp, which may be an artefact of unit curation but realistically be inconsequential.

• Andrew talked with Kevin Burrage last Friday about how stochastic systems might potentially be represented in CellML 1.2
• The audiocast of this conversation can be found at: http://www.gcast.com/u/millerak/main
• The conversation was useful but Andrew's proposal on how to represent stochastic systems in CellML 1.2 still stands as it was.
• Presently no further discussions with Kevin Burrage are scheduled.
  
• There are a number of classes of stochastic systems we could represent. Andrew's proposed approach is to allow generic representation and then identify specific classes, such as systems described by the Gillespie algorithm.
• Andrew noted that while we don't have to explicitly provide a representation of stochastic systems in CellML 1.2, we should be making 1.2 generic enough to allow for a secondary specification representing these systems.

• The CellML 1.1 tutorial currently on the site is old. There are a number of new CellML 1.1 files that could be used to create a better tutorial, such as Randall's 'buckets' model, or Jonna's published 1.1 models of cardiac systems.
• Catherine is keen to get a new CellML 1.1 tutorial up

• Is CellML currently being used in any teaching?
• Not in Auckland, but Poul is keen to look into it
• He is giving a set of lectures of physiological control systems as described using ODEs and CellML may be pretty useful for this.