R/mod_latent_variables.R

In allaway/kairos: kairos: NF Analysis Explorer

# Module UI
  
#' @title   mod_latent_variables_ui and mod_latent_variables_server
#' @description  A shiny Module.
#'
#' @param id shiny id
#' @param input internal
#' @param output internal
#' @param session internal
#'
#' @rdname mod_latent_variables
#'
#' @keywords internal
#' @export 
#' @importFrom shiny NS tagList 
#' @import ggplot2
#' @import plotly
mod_latent_variables_ui <- function(id){
  ns <- NS(id)
  
  tagList(
    h2("Latent Variable (LV) Analysis"),
    box(h4('Module Summary'), 
        p("Transfer learning techniques can leverage large, well-curated datasets such as recount2 to identify patterns of gene expression (latent variables, LVs) in large datasets and assess the expression of those genes in other, smaller datasets. 
          Many of these LVs are composed of genes that map to known signatures such as Kyoto Encyclopedia of Genes and Genomes (KEGG) or Gene Ontology (GO) terms,
          while others are uncharacterized and may allow the detection of novel and meaningful transcriptomic patterns in NF data. 
          Here, we have used this approach to reduce gene-based expression data to individual latent variables. This can provide multiple benefits—LVs can highlight differences in known biology in sets of samples, 
          they can uncover previously unknown biology, and they can reduce the impact of technical and experimental differences across multiple datasets. Specifically, we transferred a machine learning model trained on recount2 to assess LV expression in the NF1 nerve sheath tumor dataset."),
         width = 12),
    box(title = "Most Variable LVs",
        plotly::plotlyOutput(ns('top_lv')
                             # %>% shinycssloaders::withSpinner(custom.css=T) ##throws an error that looks to be css related
                             ), 
        width = 6, 
        solidHeader = T,
        status = "primary"),
    box(title = "Selected LV Expression",
        plotly::plotlyOutput(ns('individual_lv_plot')
                             # %>% shinycssloaders::withSpinner(custom.css=T) ##throws an error that looks to be css related
                             ), 
        width = 6, 
        solidHeader = T,
        status = "success"),
    box(width = 6, 
        solidHeader = T),
    box(title = "Selected LV Loadings", 
        plotly::plotlyOutput(ns('lv_loadings')
                             # %>% shinycssloaders::withSpinner(custom.css=T) ##throws an error that looks to be css related
                             ), 
        width = 6, 
        solidHeader = T,
        status = "success")
    )
}
    
# Module Server
    
#' @rdname mod_latent_variables
#' @export
#' @keywords internal
    
mod_latent_variables_server <- function(input, output, session, specimens){
  
  output$top_lv <- plotly::renderPlotly({
    
   foo <- kairos::latent_var %>% 
      dplyr::filter(specimenID %in% specimens()) %>% 
      dplyr::group_by(latent_var) %>%
      dplyr::summarize(sdev = sd(value)) %>% 
    dplyr::top_n(10, sdev) 
   
   validate(need(nrow(foo)>0, "No latent variable data found. Please modify your cohort."))
   
   foo2 <- dplyr::filter(kairos::latent_var,
                         latent_var %in% foo$latent_var) %>% 
     dplyr::mutate(latent_var = stringr::str_trunc(latent_var, 15)) %>% 
     dplyr::mutate(x = latent_var, y = value)
      
   # print(foo2)
   # 
   # ggplot(foo2) +
   #  geom_boxplot(aes(x=, y=value)) +
   #   theme_minimal() +
   #   theme(axis.text.x = element_text(angle = 45)) +
   #   labs(x = 'Latent Variable', y = "Expression")
   # 
   #  plotly::ggplotly(source = 'lv_overview', tooltip = 'x')
     create_boxplot(foo2, 
                            source_name = "mod_lv_a", 
                            color_col = NA,
                            sort = "desc") %>% 
       remove_legend() 
  })
  
  output$individual_lv_plot <- renderPlotly({
    d <- event_data("plotly_click", source = 'mod_lv_a')
    lv <- unique(d$x)
    
    validate(need(length(lv)==1, "Click a box on the left plot to examine individual latent variables."))
    
    foo <- kairos::latent_var %>% 
      dplyr::filter(specimenID %in% specimens()) %>% 
      dplyr::mutate(latent_var = stringr::str_trunc(latent_var, 15)) %>% 
      dplyr::filter(latent_var == lv)  %>% 
      dplyr::mutate(x = modelOf, y = value)
    
    create_boxplot(foo, 
                          source_name = "mod_lv_b", 
                           color_col = NA) %>% 
      remove_legend() 
    
  })

  output$lv_loadings <- renderPlotly({
    d <- event_data("plotly_click", source = 'mod_lv_a')
    lv <- unique(d$x)

    validate(need(length(lv)==1, "Click a box on the left plot to see latent variable genes."))

    foo <- kairos::mp_loading %>%
      dplyr::mutate(latent_var = stringr::str_trunc(latent_var, 15)) %>%
      dplyr::filter(latent_var == lv) %>%
      dplyr::mutate(x = hugo_gene, y = loading)

    create_barplot(foo,
                           source_name = "mod_lv_c",
                           color_col = NA,
                           sort = "desc") %>%
      remove_legend()

  })

    
  ns <- session$ns
}
    
## To be copied in the UI
# mod_latent_variables_ui("latent_variables_ui_1")
    
## To be copied in the server
# callModule(mod_latent_variables_server, "latent_variables_ui_1")