How recommended is sampling of solution space?

[varshit dusad]

Hi all,

I have a query about sampling. If I am accurate, FBA chooses one of the extreme solution points from the solution space. To explore the possibilities, one should do uniform sampling of the solution space. However, does that mean should I always sample my solution space for each simulation? I am working with a panel of 60 cancer models with ~4000 reactions each. Sampling for one simulation using gpSampler > 10  minutes at the very least. If I choose to sample the space every simulation then it would become computationally very challenging.

Can anyone please shed some light on how to judge when one should opt for sampling and when not?

Many thanks in advance!

Warm regards,

Varshit Dusad

[Jan Schellenberger]

Hi Varshit - what are you trying to do and what are your "simulations"?  Generally if you do FBA, you would get one result and you can use sampling to explore the space of equivalently optimal results.  But you only need to do this once.  

If you have 60 models and you want to sample the optimal solutions of each, then I would think you'd need to sample all 60.  You could parallelize the work of course.  

-Jan

--

---
You received this message because you are subscribed to the Google Groups "COBRA Toolbox" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cobra-toolbox+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[varshit dusad]

Hi Jan,

Thank you for replying!

Sorry for incomplete information on my part. I generally aim to do Flux balance analysis though in different conditions. If I understand correctly, for every set of constraint on a model the solution space will be different. Also, my understanding is that the optimal solution chosen by FBA may not be the best fir for all the fluxes. Therefore, sampling the solution space is more appropriate. However, doing sampling, again and again, is computationally expensive and I wanted to know if there is a smarter way to proceed about this issue.

Thank you!

Warm regards,

Varshit Dusad

To unsubscribe from this group and stop receiving emails from it, send an email to cobra-toolbo...@googlegroups.com.

[Thomas Pfau]

Hi Varshit,

One option that you could use is to do FVA instead of sampling. FVA will inform you about the flux ranges all fluxes can attain and should be much faster than sampling.

Also, my understanding is that the optimal solution chosen by FBA may not be the best for for all the fluxes.

This is not really correct.
While yes, there can be alternative optimal flux distributions in FBA, each of those flux distribution is optimal with respect to the requested objective. So it depends a lot on what you consider "the best" solution for all fluxes.

Best

Thomas

On 2018-06-07 22:18, varshit dusad wrote:

Unsubscribe

It appears that you have subscribed to commercial messages from this sender. To stop receiving such messages from this sender, please unsubscribe

For more options, visit http://secure-web.cisco.com/1NplxHN88zxbZTqIBBfpK5PhsPCOeB6igbVLwTmY2sDGBhZI6WJ5RRe5btt8EiqsoY5VTzQsnUR59GYuCIBzO4_8aKb4cQcW7iOFw2g5UYLKbfyYIV_v6VU_d9ObrOxn-ZueqZsd7TT_PSjANn6Dvndv58nZ5MPceQehd7sbkVn0er2hTEFiP7YrJ95A4Ulhe-vjphUmoDBW8EmDq4SuyTksVrrJos3AfTyxTKnSLtWFttuSNvY22-oJM5Crk1GYRk2aSBAxIi3W6LttBTsD47Xekj1r-qA6qH_xQEawUyQHDgtnC99rfZP7OaTAniLIOEILIZpl98FBjz8F0ZptUkDi9sUJGWb2wDf5LrucaRFFP-99LhXA_u0taqtWazZMxiSzy2mbBS-dd8Aw_VQ-_bg/l34%3Ahttps%3A%2F%2Fgroups.google.com%2Fd%2Foptoute.

-- 
Université du Luxembourg
Faculté des Sciences, de la Technologie et de la Communication
Campus Belval, Biotech II 115
6 avenue du Swing
L-4367 Belvaux
Tel: (+352) 46 66 44 5309
Email: thoma...@uni.lu