RMI2_Random_Forest.Rmd

params:
  Cells:
    label: 'Cell type'
    value: 'Lymphocytes'
    input: select
    choices: ['Lymphocytes']
    
  Activations:
    label: 'Activation'
    value: 'Resting'
    input: select
    choices: ['Resting', 'Activated']

  Score:
    label: 'mRNA degradation index'
    value: 'Total Degradation'
    input: select
    choices: ['Total Degradation', 'Translation Independent Degradation', 'Translation Dependent Degradation']

title: "RMI2 Random Forest for Translation Dependent Degradation \n`r params$Cells` `r params$Activations`\n`r params$Score`"

output: 
  pdf_document:
    toc: true
    toc_depth: 4

header-includes:
  - \usepackage{titling}
  - \usepackage{fancyhdr}
  - \pagestyle{fancy}
  - \fancyhf{}
  - \lhead{`r format(Sys.time(), '%Y/%m/%d %H:%M:%S')` \newline `r params$Cells` `r params$Activations`}
  - \chead{`r params$Score`}
  - \rhead{
      \includegraphics[width=2cm]{Logos/Logo-ens.jpg}
      \includegraphics[width=1cm]{Logos/Logo-LBMC.jpg}
      \includegraphics[width=1cm]{Logos/logo-rmi2-lab-version-2.png}
      }
  - \cfoot{\thepage}
  - \lfoot{Written by Labaronne Emmanuel and Cluet David \newline current version 2023/12/06}
  - \renewcommand{\footrulewidth}{0.4pt}
  - \pretitle{\begin{center}
    \includegraphics[width=2cm,height=2cm]{Logos/logo-rmi2-lab-version-2.png}\LARGE\\}
  - \posttitle{\end{center}}

knitr::opts_chunk$set(echo = TRUE)

\newpage

Column name reconstruction

# Generate the score column name from the explicit GUI elements

if (params$Cells == 'Lymphocytes') {
  cells <- 'Lympho_'
}

if (params$Score == 'Total Degradation') {
  score <- 'DegFold.'
  suffix <- '.Trip.Ref_Trip_0h.3h'
} else if (params$Score == 'Translation Independent Degradation') {
  score <- 'Abs.NonTDD..'
  suffix <- '.Trip_CHX.Ref_Trip_0h.3h'
} else if (params$Score == 'Translation Dependent Degradation') {
  score <- 'Abs.TDD..'
  suffix <- '.Trip_CHX.Ref_Trip_0h.3h'  
}

# Concatenate all sub-parameters to generate the true column name
x <- paste0(score, cells, params$Activations, suffix)
x_explicit <- paste0(params$Cells, '_', params$Activations, '_', params$Score)

# The following code is an optimisation of Emmanuel Labaronne original script

\newpage

Loading of the packages and the source functions

# Load functions
source("src/randomForest.R")

\newpage

Fine grain tuning of the Random Forest model

# Perform the random forest
model <- makeRandomForest(x = x,
                          explicit_x = x_explicit)

\newpage

Access to all generated models

# Access to all models
model[["model"]]

\newpage

Access to the best model

# Get the fine tuned model
model[["model"]]$finalModel

\newpage

Model evaluation on Training and Validation set

suppressWarnings(grid.arrange(model[["training_set"]], model[["valid_set"]], ncol = 1))

\newpage

Importance of the chosen parameters inside the model

suppressWarnings(model[["ImpVar"]])

\newpage

SHAP summary

suppressWarnings(model[["shap_summary"]])

\newpage

SHAP dependences

suppressWarnings(gridExtra::grid.arrange(grobs = model[["shap_dep1"]], ncol = 3))

\newpage

suppressWarnings(gridExtra::grid.arrange(grobs = model[["shap_dep2"]], ncol = 3))