Skip to content

B2-Bayesian-for-Biology/matlab-ode-inference

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

matlab-ode logo

matlab-ode-inference: ODE based Bayesian modeling of dynamics with Matlab

matlab-ode-inference is a repository to study Biologically inspired dynamical systems with mechanistic models (such as ODEs). Inferring parameters of a mechanistic models involves answering a lot of questions:

  • How can we incorporate prior information about parameters and models into our modeling framework?
  • How does the choice of the model affect the estimates of the inferred parameters?
  • How can we use a Bayesian approach (for example, MCMC type samplers) to model these systems?

We want to compare different methods and techniques on simple in-silico of experimental datasets.

Installation 📥

First clone the mcmcstat package and add it to Matlab's path.

git clone git@github.com:mjlaine/mcmcstat.git
git clone git@github.com:B2-Bayesian-for-Biology/matlab-ode-inference.git

Getting started

Once you have cloned the repository you can look at a simple example of modeling a growth-curve of cells driven by resources, modeled by an ODE.

Simple demonstration

demo-example

demo-example-posterior

Comparing between adaptation steps:

How to tune the adaptation step of a Delayed rejection? TO DO

Data transformation/ reparametrization:

  • Should the data be log transformed?
  • Should some of the parameters be sampled in the log scale?

FAQ

  • Language and implementation: MAtlab -- mcmcstat -- very slow! But transparent!
  • ODE solver: ode45.
  • Error Model: inv-gamma; assumped independence of points, multiplicative likelihood, compared log(data) with log(model)
  • Convergence tests: Gelman rubin, Autocorrelation of chains, geweke.
  • Post processing: burn = 3000 -- need burning for DRAM-MCMC, need atleast 10 chain thinning for.

Review:

Here's the convention we used (for now).

  • Initial condition fixed from data for P0 and varied for N0
  • flags: Logtransformed = 1 means log(Data) and log(model) compared.
  • flags: Logprior = 1, LogNormal Priors used for N0 and Qs
  • You can find chains saved in results
  • You can find plots saved in figures

Contributions.

  • Current contributions by Raunak Dey at UMD.
  • Part of the big-project led by David Talmy at UTK.
  • Please feel free to reach out to me at my email

About

Matlab based non-particle samplers to performed ODE based Bayesian parametric inference

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages