shinyGMM implements a collection of R code written during prototyping for nQuire into a interactive shiny application. It implements inference of Gaussian Mixture Models (GMM) using either an Expectation-Maximization (EM) algorithm or a Gibbs sampler. Some of the code was adapted from Matthew Stephens’ fiveMinuteStats.
The primary purpose of this project was to be able to play around with simulations of GMMs
and how the different inference methods deal with changing parameterization. However, it also
implements an interface with .bin files generated by nQuire, so real base frequency data
can be loaded. Still, it is very experimental and by no means intended for production data analysis. It
does not use the implementation of the EM algorithm used in nQuire, but its R based prototype.
Clone the repository:
git clone https://github.com/clwgg/shinyGMMCompile the C plugin for nQuire .bin file integration and install all R
packages needed (these are: shiny, magrittr, dplyr, tidyr, ggplot2, cowplot, gtools):
cd shinyGMM
make lib
make packagesTo run the application, use:
make runAt the end of the start-up procedure, there will be a link to the localhost port where the application is listening. This should be opened in a suitable web browser.