In dkaschek/dMod: Dynamic Modeling and Parameter Estimation in ODE Models

knitr::opts_chunk$set(echo = TRUE, fig.width = 7, fig.height = 3.5)

You might want to interactively browse through your results and share this view with your colleagues. A nice way to do so is via a shiny web app, see https://shiny.rstudio.com/.

Check out a customized shiny app for showing your modelling results obtained with dMod at http://fermi.fdm.uni-freiburg.de:3838/mfehling/dModtoShinyPub/.

Contents of the app

The main input to the app is a list of objects (a list of fits generated with mstrust, data, the prediction function, parameter profiles, list of reactions) or a dMod.frame. Implemented features include the standard dMod plots:

Plot Prediction: based on plotCombined to plot the prediction and data for chosen time interval, conditions and states (subsetting tab) for different options (general tab, e.g. log-axis). By default only the best/global optimum is plotted but you can choose more optima on the next tab Optima Selection.
Optima selection: calls plotValues and optionally plotPars. Allows to choose several optima to be shown in other tabs, e.g. Plot Prediction and Residuals and to export the parameter values from one optimum to the Parameter Tuning tab.
Parameter Tuning: see how the prediction changes when changing parameter values. All parameter values can be changed and the new prediction is shown as solid line, while the original prediction is shown as dashed line. Optionally, the change of the fluxes can be viewed instead.
Run on Profiles: includes several options regarding the parameter profiles. Calls plotProfiles (click Plot profiles) or plotPath (switch to Path tab above the plot), or change the parameters along one choosen profile and plot the corresponding prediction (see Choose profile in the left panel).
More Plots: contains two options:
- Fluxes: calls plotFluxes, choose from fluxes predefined by the model and/or type your own flux-equations
- Residuals: calls plotResiduals, choose from several options to summarise the residuals
View Reactions: print model reactions in tabular format
Optinal link to paper if results are published

Run the app locally

# in case you don't have those packages installed already, install:
install.packages("shiny")
install.packages("shinycssloaders")
install.packages("shinyjs")
install.packages("shinydashboard")

The app (app.R) and one example model can be found in the inst/examples/dModtoShiny folder. Run the shiny app locally on your machine:

library(dMod)
library(shiny)

setwd(paste0(system.file(package = "dMod"),'/examples/dModtoShiny/'))

runApp()

The app reads in models from all subfolders within the app folder dModtoShiny (in this case there is only one model CCD4). The app might not run using the input of the existing folder since it contains compiled code. In this case you need rebuild the contents of the CCD4 folder:

library(dMod)

setwd(paste0(system.file(package = "dMod"),'/examples/dModtoShiny/CCD4'))
source(paste0(system.file(package = "dMod"),'/examples/example_CCD4/setup.R'))
myprofiles <- readRDS(paste0(system.file(package = "dMod"),'/examples/example_CCD4/profiles.Rds'))
testtrust <- readRDS(paste0(system.file(package = "dMod"),'/examples/example_CCD4/mstrust_CCD4.Rds'))

saveShiny <- function(reactions, 
                      x, 
                      parameters, 
                      fixed = NULL, 
                      data, 
                      profiles, 
                      pubref="none", 
                      errmodel = NULL){
  save(reactions, x, fixed, data, parameters, profiles, pubref, file = "input_shiny.RData")
}

saveShiny(x = g*x*p, parameters = testtrust, data = data, reactions = f, profiles = myprofiles, pubref = "https://www.ncbi.nlm.nih.gov/pubmed/27811075")

setwd(paste0(system.file(package = "dMod"),'/examples/dModtoShiny/'))

runApp()

Add more models by adding one subfolder per model. The necessary input needs to be stored in a file called input_shiny.RData which can be generated by the saveShiny function shown above. In addition to the .RData file, corresponding .so files need to be copied to the model subfolder. Easier exports are work in progress (see input_shiny.R for a preliminary export using the dMod.frame). Remark: it is currently not possible to add different setups of the same model (e.g. different parameterization) in one folder, but you can add a folder for each setup.

To customize the app, modify app.R.