MATLAB Answers

0

Is there a good tutorial for incorporating some of the MCMC code into a custom task in Simbiology?

Asked by Andrew Heitman on 30 May 2019
Latest activity Answered by Andrew Heitman on 4 Jun 2019
I Am looking to add MCMC as a custom task and have no idea where to start in creating this - I Assume a custom task -but I am looking for a guid in how to even start writing this code.
Any direction is helpful.

  0 Comments

Sign in to comment.

Products

Communities

More Answers in the  SimBiology Community

2 Answers

Answer by Fulden Buyukozturk on 4 Jun 2019
 Accepted Answer

Hi Andrew,
You can implement this, or custom analyses in general, in SimBiology either by creating a custom task on SimBiology desktop as you've suggested or by using a MATLAB script. I would suggest using a MATLAB script at least initially because it will be easier to debug your code.
To help start with this, you can configure a task and then check out the MATLAB/SimBiology code behind that task in SImBiology. This could serve as a starting point for your custom analysis. You can see the code by clicking on Task Code -> View Task Code on any built-in task on task editor. Please see below.
Picture1.png
For example, to run a simple simulation, the programmatic workflow that you can run directly via a MATLAB script could look something like the following (you can run this example too since i'm using a shipped project):
% Load model from SimBiology project called AntibacterialPKPD
project = sbioloadproject('AntibacterialPKPD','m1');
model = project.m1
% Get model simulation settings and set stop time to 336
settings = getconfigset(model);
settings.StopTime = 336;
% Get the dosing schedule from the model
dose = getdose(model, '500 mg tid')
% Simulate with the dose
simdata = sbiosimulate(model, settings, dose);
% Plot all states
sbioplot(simdata)
For implementing MCMC specifically, I suggest you take a look at simfunction and Hamiltonian Monte Carlo sampler from the Statistics and Machine Learning Toolbox.
Does this help?
Fulden

  0 Comments

Sign in to comment.


Answer by Andrew Heitman on 4 Jun 2019

Yes, this is a great place to start. Thanks!

  0 Comments

Sign in to comment.