Unusually slow Parameter Estimation

[Theo Redfern]

Dear all,

I hope this finds you well!

I've been having trouble with my model recently in COPASI. Parameter Estimation is very slow all of a sudden and there are certain parameter combinations for which the time-course simulations are not able to be solved. 

I thought failing to solve an ODE can happen when some reactions are too quick to be solved with a given timestep. But changing the timestep and interval etc seem to have little if any effect. Additionally it appears to happen even when I reduce rate constants, not increase them up (hence reactions should be slower if anything).

Additionally, I'm having trouble doing Parameter Estimation now. Previously I was able to calculate around 90-100 Evaluations/second on my laptop but it is now at 1.26 Evaluations/second...

Could there be some mistake or error somewhere in my model causing this? Or could having many plots and global quantities be causing an issue? Are there any things that are generally recommended for speeding up computation time in COPASI?

Thank you for any help you can offer, (my COPASI file is attached, though without the Experimental data in Parameter Estimation)

I hope the questions aren't too much too!

I really appreciate the service you all provide with this User Forum and COPASI itself is a really impressive software!

Best wishes,

Theo

[Hoops, Stefan (sh9cq)]

Hello Theo,

You may want to look:

1) Parameter ranges 10^-9 to 10^9 are 18 orders of magnitude. This will
create often very stiff problems
2) The Max Internal Steps of LSODA is 10 million that means the
integrator will take very long before it fails on a stiff problem

Both problems by itself may create long running time courses even for
medium size models

Thanks,
Stefan

--
Stefan Hoops, Ph.D.
Research Associate Professor
Biocomplexity Institute & Initiative
University of Virginia
995 Research Park Boulevard
Charlottesville, VA 22911

Phone: +1 540 570 1301
Email: sho...@virginia.edu

[Theo Redfern]

Hey Stefan,

Ah great thank you. I think I will do some more research into how this all works then to appreciate your answer and how the solvers work. Do you have any particular recommendations literature-wise etc?

Thanks,

Theo

[Mendes,Pedro]

Theo,

your best option is to restrict the parameter range that you are
searching for. rather than using 1e-9 to 1e9 try a smaller range, eg
1e-8 - 1e3 or even 1e-6 - 1e3

Then if the fits make a parameter hit the boundary, you could expand
that boundary for that parameter (just that one). This should improve
your run times.

Pedro

> Phone: +1 540 570 1301 <tel:(540)%20570-1301>
> Email: sho...@virginia.edu
>
> --
> You received this message because you are subscribed to the Google
> Groups "COPASI User Forum" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to copasi-user-fo...@googlegroups.com
> <mailto:copasi-user-fo...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/copasi-user-forum/76b06cb7-5e89-40ba-ac1e-3f44f05421dcn%40googlegroups.com
> <https://groups.google.com/d/msgid/copasi-user-forum/76b06cb7-5e89-40ba-ac1e-3f44f05421dcn%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Pedro Mendes, PhD
Professor and Director
Richard D. Berlin Center for Cell Analysis and Modeling
University of Connecticut School of Medicine
group website: http://www.comp-sys-bio.org

[Theo Redfern]

Dear Dr Mendes,

Excellent thank you, I will try this!

Best,

Theo