New features since the previous stable release:

Graphical User Interface (CopasiUI):

• Added plots and reports for parameter scans to the output assistant.
• Only applicable plots and reports are shown in the output assistant by default.

Simulation Engine:

• Time Scale Separation Analysis can now handle multi compartment models.
• The Time Scale Separation Analysis can now be applied to the full as well as the reduced model.
• It is now possible to calculate the sensitivities of Eigen values.
• Add kinetic function Bi (irreversible) to the builtin functions.

Download available from Copasi.org.

There’s still no ability to copy reactions, which would be nice for making models of systems like MAPK cascades with sequential reactions that are very similar to the previous reaction.  It would also be handy to be able to highlight and copy the data in the various matrices that are produced into the clipboard.  This which would make it quicker to move data around software like Excel without having to generate and import reports.  The software does generate report ouputs though, so this is just an annoyance rather than a problem.

Copasi still lacks the capability to create 3D surface plots for multiple parameter scans, or find bifurcation points like xppaut, which considering this is one of xppaut’s primary strengths you would think the developers would incorporate into Copasi.  As I said above though, Copasi can output reports, so you can attempt to wrestle the raw data through the python bindings and plot using matplotlib or just dump it into sigmaplot or origin.

There’s still no statistical analysis for stochastic simulations.  It would be useful to generate mean and standard deviation values for repeats of stochastic simulations, as the raw data outputted in the reports is virtually useless, unless you can write your own scrupt to sort it and calculate the statistics.  It’s impossible to do by hand.  As stochastic simulations require multiple trajectories that are then compiled into mean and SD for things like signal:noise ratios this is a major pain.  The developers did help me a while back though with a python script to do this, which was very much appreciated.  This was about 2 years ago and it would have been nice to see this implemented in the software.

You still can’t program in custom rate laws and have the save if they’re not assigned to a reaction when you exit the software.  This is annoying as I have different strength promoters that I was representing with different rate laws and switching between them between simulations.  Now I have to have multiple models or code them each time I load the software.  I ended up working around this though by assigning global quantities to parameters in a generic rate law.  It’s a bit more elegant than having multiple unused rate laws in your model file I suppose.  I can also then manipulate these global quantities with other rate laws and output them with the data so it does provide a lot of flexibility.

All things considered, Copasi still remains the best of the GUI based modelling tools with the amount of optimization algorithms and solvers, and it’s free, so you can’t really complain at the glacial development.  The average Joe biologist (me) just couldn’t code this stuff in Matlab or Fortran.  The Python bindings provide the potential to couple it up with xppy, to use Copasi’s solvers and optmization routines with Python as a hub and augment it with functions from PySces and xppaut.  I hope to look more into using Python as a central controller for these tools once I get some time.

Graphical User Interface (CopasiUI)

• Enhanced parsing of mathematical expression to allow the comparison operator , >=, and == as well as the Boolean operators || and &&.
• Allowed calculation results to be selected as constraints for optimization and parameter estimation.
• Improved display of mass conservation results. The moieties are now displayed in amount if the user is in the concentration framework.
• Added Update Model button to the optimization result widget
• Notes in text or XHTML format are now available for compartments, species, reactions, global quantities, events, and kinetic functions.
• Support links in XHTML notes.
• Support of render informations in graphical model layouts.

Simulation Engine

• The calculation of statistics is now optional for optimization and parameter estimation tasks.
• Start values can now be randomized automatically for optimization and parameter estimation tasks.
• Enhanced MCA algorithm performance by applying a new selection criterion before each internal step.
• Added an stochastic algorithm (Adaptive SSA/Tau-Leap) which dynamically partitions the model into parts simulated by the direct algorithm and the Tau-Leap algorithm.

SBML

• Added support for constant conversion factors in Level 3.
• Import SBML Level 3 Version 1.
• Added SBML notes support for compartments, species, reactions, global quantities, events, and kinetic functions.
• Added MIRIAM annotation support for events.
• Support for the SBML Render Extension.

Copasi is available from www.copasi.org

I stumbled over a couple of blogs explaining how it’s possible to embed Python code into LaTeX, using python.sty programmed by Martin R. Ehmsen (another good tutorial is here as well). The author’s website is producing 404 errors and there is no download for python.sty. I found a copy on the web and have it available here, purely because I can’t link to it from the author.

There is some example code for embedding Python script, from over at the texexample.net blog:
%& -shell-escape
\documentclass{article}
\usepackage{python}
\begin{document}

Say hello Python:
\begin{python}%
print r"Hello \LaTeX!"
\end{python}%
\end{document}


You have to remember to compile your LaTeX script with –shell-escape option. The post also continues to do some plotting with PyX.

For embedding Cairoplot in LaTeX there is some example code at Michael Wayne Goodman’s blog (goodmami.org)

Some example code for a dot plot from the Cairoplot web page is below:
dot_line_plot (name,
data,
width,
height,
background = None,
border = 0,
axis = False,
grid = False,
h_legend = None,
v_legend = None,
h_bounds = None,
v_bounds = None)

name – Name of the desired output file, no need to input the .svg as it will be added at runtim;
data – The list, list of lists or dictionary holding the data to be plotted;
width, height – Dimensions of the output image;
background – A 3 element tuple representing the rgb color expected for the background. If left None, a gray to white gradient will be generated;
border – Distance in pixels of a square border into which the graphics will be drawn;
axis – Whether or not the axis are to be drawn;
grid – Whether or not the gris is to be drawn;
h_legend, v_legend – lists of strings containing the horizontal and vertical legends for the axis;
h_bounds, v_bounds – tuples containing the lower and upper value bounds for the data to be plotted.

They use the following example:
teste_data_2 = {"john" : [10, 10, 10, 10, 30], "mary" : [0, 0, 3, 5, 15], "philip" : [13, 33, 11, 25, 2]}
teste_h_legend = ["jan/2008", "feb/2008", "mar/2008", "apr/2008", "may/2008"]
CairoPlot.dot_line_plot('teste2', teste_data_2, 400, 300, h_legend = teste_h_legend, axis = True, grid = True)

to plot an example line type plot with multiple lines, such as used quite often in biology publications.

I recreate the code here to keep a record for myself, the original should be viewed on the Cairoplot page with the output images.

But now you too can have opengl rendered graphs in your thesis! :p

\DTLloaddb{my-table}{./path-to-your-csv-file/my-table.csv}
\DTLdisplaydb{my-table}


will display your csv in a default table in your document.

Simple as.

A comment by Herbert Sauro below also points out that it would be interesting to see if it still links with AUTO, as the most popular functionality in the desktop edition of XPPAUT is bifurcation analysis.

XPP for iPad is interesting though as Universities are considering supplying undergrad programs with tablets, pre-loading them with XPP would be ideal for computer science and systems biology students.

Available for free in your local iTunes store.

One of the things I liked about the Windows versions of Adobe and Foxit readers is the click and drag scroll feature. You can use the mouse to slowly pan down the document as you read. I find this easier than having to move my hands to the keyboard arrows because I’m terminally lazy and my eyes lose the line I’m on as I scroll. Unfortunately this feature doesn’t exist in the stock Evince installation. However! I stumbled over this blog post over at TechiTalk.info that details a patched version of Evince with the drag scroll option by IRIE Shinsuke.

Add the PPA with: sudo add-apt-repository ppa:irie/evince

Once installed just add the hand drag icon to the tool bar and have it selectable as your scroll method. More detailed info is available over at TechiTalk blog or the Shinsuke’s launchpad page if required.

Happy days.

For the non-computer scientist / mathematicians, XPPAUT can provide some very fast results and give your matlab jedi colleagues a run for their money. There are a number of ODE solvers, including stochastic algorithms, and more advanced features include parameter fitting of experimental data using the Levenberg–Marquardt algorithm and even animations of your systems. Combined with Copasi’s range of parameter scanning and fitting features, these two tools are potentially very empowering for the biologists in systems biology. Personally, I think xppaut is much more diverse than Copasi, particularly the model script being able to input mathematical formulae, and I would have recommended it to our 1st year doctoral training program. The ability to visualize fixed points and watch the eigenvalues changing with the stability helped to understand this as a non-mathematician.

I’m off to find more Fitzhugh-Nagumo fixed points…

Update: For those interested in the software, there is an accompanying book: Simulating, Analyzing, and Animating Dynamical Systems by Bard Ermentrout which is really easy to read and goes through all the functions with worked examples and is clearly illustrated.

Much recommended if you’re working the graveyard shift.

When I started my Ph.D many of the Python based systems biology tools were still emerging and I struggled to get them running, so I always defaulted back to Copasi or Matlab. There also weren’t many available other than scipy and numpy and the ODE integrators just didn’t compete with Matlab’s ODE45 (or I couldn’t seem to find ODE45 for Python). More recently however a number of tools have continued to develop and look very interesting.

I started up with the standard installation of python 2.6 in Ubuntu and added SciPy and NumPy from the repositories, with matplotlib. I then installed PySCes, Stompy, PyDSTools, and the Copasi language bindings.

PySces (Brett G. Olivier et al, Stellenbosch University) is a general toolset for all things to do with simulating cellular systems. It’s now compatible with SBML so I could import all the models I had built up in Copasi. (The process of exporting from Copasi to SBML and into PySCes however loses the names of all the species which is annoying in larger models). PySCes can do timecourse simulations with LSODA and CVODE, calculate steady state with HYBRD and NLEQ2, do MCA, structural analysis, bifurcation analysis with PITCON, and n dimensional parameter scans.

Stompy (Timo Maarleveld and B. G. Olivier, also Stellenbosch University) is in extension to PySCes that adds stochastic modelling to the above tools. It includes Direct, next reaction, and Tau leaping algorithms. Stompy can also plot the simulation output and do data analysis such as propensities, distributions, waiting times etc. The website includes a number of clear tutorials and is quite simple to use.

PyDSTool (Prof. Rob Clewley lab, Georgia State University) is another modelling toolbox with ODE/DAE/discrete map simulation tools, bifurcation analysis, symbolic expression utilities, and is compatible with SciPy and the SBML file format.

I’m not sure how much overlap there is between PySCes and PyDSTool in terms of functionality. I have not used either in great depth yet. PyDSTool has a nice tutorial for finding and investigating a bifurcation point.

Both tools appear to be being integrated into the graphical application Tinkercell (which also has a number of handy Python scripts for interrogating models of biological systems).