LSODA error

[Lu Li]

Hi,

I am using deterministic LSODA method to simulate some kinase response to high frequency calcium spikes (calcium inputs as event).
I got the following error message during calcium inputs:
"

>EXCEPTION 2017-09-05 16:43:28<

CTrajectoryMethod (6): Deterministic integration failed. LSODA reported:

DLSODAR- One or more components of g has a root

too near to the initial point.

DLSODAR- One or more components of g has a root

too near to the initial point.

DLSODAR- One or more components of g has a root

too near to the initial point.

DLSODAR- At T (=R1) and step size H (=R2), the

corrector convergence failed repeatedly

or with ABS(H) = HMIN

In above message, R1 = '10.241', R2 = '2.29127e-11'

DLSODAR- One or more components of g has a root

too near to the initial point.

DLSODAR- At T (=R1) and step size H (=R2), the

corrector convergence failed repeatedly

or with ABS(H) = HMIN

In above message, R1 = '10.241', R2 = '2.29127e-11'

Please see result for indications of numerical instability

"

I try to adjust relative/absolute tolerance or/and increase MAX internal steps, and sometimes I can solve this problem but some times I can not.

Could you please explain this error message and give me some suggestions?

Thank you in advance

Lu

[Stefan Hoops]

Hello Lu,

The crucial error is the last on reported:

DLSODAR- At T (=R1) and step size H (=R2), the
corrector convergence failed repeatedly
or with ABS(H) = HMIN
In above message, R1 = '10.241', R2 = '2.29127e-11'

Which tells you that at Time = 10.241 the integrator is not able to
achieve the desired tolerance (absolute or relative) with the minimal
allowed step size = 2.29127e-11, which is internally determined.

Without looking at your model I can suggest that you should check
whether the system you are integrating has a singularity near time
10.241, i.e., one of your variables increases exponentially.

Thanks,
Stefan

--
Stefan Hoops, Ph.D.
Research Assistant Professor
Director, Biochemical Networks Modeling Group
Faculty of Health Sciences (FHS)
Biocomplexity Institute of Virginia Tech
1015 Life Science Circle (0477)
Blacksburg, Va 24061, USA

Phone: (540) 231-1799
Fax: (540) 231-2606
Email: sho...@vt.edu

[Lu Li]

Hi Stefan,

Thank you very much!
Indeed, I am running a train of calcium inputs at 50Hz for 100 spikes. I was running the time course with a 1-millisecond interval, however it seemed not sufficient to get calcium concentration down after each spike, therefore I had the singularity issue.

It seems using a much smaller time interval (0.0001s) helps to run through the simulation. However, it makes the whole simulation extremely slow, especially the equilibrium part before calcium spikes start.

This brings me to two questions:
First of all, I always thought LSODA can adjust the simulation interval size and refine it based on processes, so the part without calcium spikes should run faster regardless of interval defined in time course setting. Am I right?

Secondly, (if LSODA can not achieve above for whatever reason), is there a way that I can manually force Copasi run a time course with different time intervals, so that at the "equilibrium" run, I can have a 1-second spacing whereas when calcium spikes start, I can fine-tune to 0.0001s interval?

Thank you so much for your help
Best regards
Lu

[Stefan Hoops]

Hello Lu,

The step size or number of outputs do only indirectly influence LSODA's
internal steps. The internal step size is determined by error estimates.
What can happen due to a large number of interpolations is that LSODA
detects that it need to reevaluate the Jacobian sooner that if left
alone. Thus you see different results.

Since you want to start in equilibrium you may want to check:
Start in Steady State

Thanks,
Stefan


On Wed, 6 Sep 2017 06:32:32 -0700 (PDT)

> > Email: sho...@vt.edu <javascript:>