Merge pull request #25 from allaway/fix-lv

fixed a couple of bugs and cleaned ui

NAMESPACE

@@ -7,6 +7,8 @@ export(mod_cohort_page_server)
 export(mod_cohort_page_ui)
 export(mod_docs_page_server)
 export(mod_docs_page_ui)
+export(mod_drug_screening_server)
+export(mod_drug_screening_ui)
 export(mod_explore_page_server)
 export(mod_explore_page_ui)
 export(mod_gene_variant_server)

R/mod_cohort_page.R

@@ -22,15 +22,21 @@ mod_cohort_page_ui <- function(id){
       dashboardHeader(disable = T),
       dashboardSidebar(
           p("Globally remove or add groups with these selectors:"),
-          checkboxGroupInput(ns('isCellLine'), label = "Is Cell Line", choices = unique(cohort$isCellLine), 
-                             selected = unique(cohort$isCellLine)),
           checkboxGroupInput(ns("tumorType"), label = "Tumor Type", choices = unique(cohort$tumorType),
                               selected =  unique(cohort$tumorType)),
+          checkboxGroupInput(ns('isCellLine'), label = "Is Cell Line", choices = unique(cohort$isCellLine), 
+                             selected = unique(cohort$isCellLine)),
           checkboxGroupInput(ns("species"), label = "Species", choices = unique(cohort$species), 
                               selected = unique(cohort$species)),
           selectizeInput(ns("studyName"), label = "Study Name", choices = unique(cohort$studyName),
                           selected = unique(cohort$studyName), multiple = T)),
       dashboardBody(
+        h2("Select a Cohort"),
+        box(p("The first step to any kairos analysis is building a cohort. The selector on the left allows you 
+              to add or remove biological specimens based on metadata such as tumor type, species, and the study it comes from.
+              A graphical summary (below) indicates the analyses that are available for the cohort you've selected, and below that,
+              a downloadable table shows which specimens you've selected."),
+            width = 12),
         box(width = 12,
             solidHeader = T,
             status = "primary",

R/mod_drug_screening.R

@@ -0,0 +1,41 @@
+# Module UI
+
+#' @title   mod_drug_screening_ui and mod_drug_screening_server
+#' @description  A shiny Module.
+#'
+#' @param id shiny id
+#' @param input internal
+#' @param output internal
+#' @param session internal
+#'
+#' @rdname mod_drug_screening
+#'
+#' @keywords internal
+#' @export 
+#' @importFrom shiny NS tagList 
+mod_drug_screening_ui <- function(id){
+  ns <- NS(id)
+  tagList(
+    h2("Drug Screening Analysis"),
+    box(h4('Module Summary'), 
+        p("Altera inimicus sed no. Congue consul eripuit est in. Ut blandit interesset mea, hinc congue quo ex, qui ad laudem volumus. Ius at everti torquatos. Eum aeque verear ei, postea sensibus te his. Quo viderer epicuri postulant cu, essent appareat efficiantur nec eu. Libris lobortis vix no. Vim at elit quas ullum. Elaboraret dissentiet ius ei, vix verterem scriptorem intellegebat in, vix case inimicus ne. Ea mea tale clita. Ne oblique invenire ius, ubique laboramus est an, usu ad oblique tractatos maluisset. Vide facilisis definitiones ei vim, in viris salutatus philosophia sit."),
+        width = 12)
+  )
+}
+
+# Module Server
+
+#' @rdname mod_drug_screening
+#' @export
+#' @keywords internal
+
+mod_drug_screening_server <- function(input, output, session, specimens){
+  ns <- session$ns
+}
+
+## To be copied in the UI
+# mod_drug_screening_ui("drug_screening_ui_1")
+
+## To be copied in the server
+# callModule(mod_drug_screening_server, "drug_screening_ui_1")
+

R/mod_explore_page.R

@@ -39,7 +39,12 @@ mod_explore_page_ui <- function(id){
                    menuSubItem(
                      "Latent Variables",
                      tabName = "latent_variables",
-                     icon = icon("cog"))))),
+                     icon = icon("cog")),
+                   menuSubItem(
+                     "in vitro Drug Screening",
+                     tabName = "drug_screening",
+                     icon = icon("cog")
+                   )))),
         dashboardBody(
           tagList(
             tabItems(
@@ -57,6 +62,10 @@ mod_explore_page_ui <- function(id){
               tabItem(
                 tabName = "latent_variables",
                 mod_latent_variables_ui(ns("latent_variables_ui_1"))
+              ),
+              tabItem(
+                tabName = "drug_screening",
+                mod_drug_screening_ui(ns("drug_screening_ui_1"))
               )
               ))
         )
@@ -76,6 +85,7 @@ mod_explore_page_server <- function(input, output, session, specimens){
   callModule(mod_gene_variant_server, "gene_variant_ui", specimens)
   callModule(mod_latent_variables_server, "latent_variables_ui_1", specimens)  
   callModule(mod_immune_signatures_server, "immune_signatures_ui_1", specimens)
+  callModule(mod_drug_screening_server, "drug_screening_ui_1", specimens)
 }
 
 ## To be copied in the UI

R/mod_gene_var.R

@@ -18,6 +18,11 @@ mod_gene_variant_ui <- function(id){
 
     tagList(
 
+      h2("Gene Mutation Profiles"),
+      box(h4('Module Summary'), 
+          p("Altera inimicus sed no. Congue consul eripuit est in. Ut blandit interesset mea, hinc congue quo ex, qui ad laudem volumus. Ius at everti torquatos. Eum aeque verear ei, postea sensibus te his. Quo viderer epicuri postulant cu, essent appareat efficiantur nec eu. Libris lobortis vix no. Vim at elit quas ullum. Elaboraret dissentiet ius ei, vix verterem scriptorem intellegebat in, vix case inimicus ne. Ea mea tale clita. Ne oblique invenire ius, ubique laboramus est an, usu ad oblique tractatos maluisset. Vide facilisis definitiones ei vim, in viris salutatus philosophia sit."),
+ width = 12),
+
       # selectizeInput(ns("studyName"), label = "Study Name", choices = unique(kairos::exome_data$study),
       #                selected = unique(kairos::exome_data$study), multiple = T),
       # 

R/mod_immune_signatures.R

@@ -71,15 +71,28 @@ mod_immune_signatures_server <- function(input, output, session, specimens){
 
     foo <- kairos::immune_predictions %>% 
       dplyr::filter(specimenID %in% specimens() &
-                      method == 'mcp_counter') 
+                      method == 'mcp_counter') %>% 
+      mutate(score = log10(score+0.01)) #this is a temporary fix but not efficient...data should be transformed
 
     ggplot(foo) +
-      ggridges::geom_density_ridges(aes(x = log10(0.01+score), y=cell_type, fill = cell_type)) +
+      ggridges::geom_density_ridges(aes(x = score, y=cell_type, fill = cell_type)) +
       facet_grid(cols = vars(tumorType)) +
       theme_minimal() + 
       theme(legend.position = 'none',
             text  = element_text(size = 12)) + 
-      labs(, x = "Score (log10)", y = 'Cell Type')
+      labs(x = "log10(Score)", y = 'Cell Type')
+    # 
+    # foo <- kairos::immune_predictions %>% 
+    #   dplyr::filter(specimenID %in% specimens() &
+    #                   method == 'mcp_counter') 
+    # 
+    # ggplot(foo) +
+    #   ggridges::geom_density_ridges(aes(x = log10(0.01+score), y=cell_type, fill = cell_type)) +
+    #   facet_grid(cols = vars(tumorType)) +
+    #   theme_minimal() + 
+    #   theme(legend.position = 'none',
+    #         text  = element_text(size = 12)) + 
+    #   labs(, x = "Score (log10)", y = 'Cell Type')
 
   })
 

R/mod_latent_variables.R

@@ -19,10 +19,14 @@ mod_latent_variables_ui <- function(id){
   ns <- NS(id)
 
   tagList(
-    box(title="placeholder box for latent variable explanation",
-        width = 12,
-        solidHeader = T,
-        status = "primary"),
+    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
@@ -95,6 +99,7 @@ mod_latent_variables_server <- function(input, output, session, specimens){
     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)

data-raw/analyses.R

@@ -13,7 +13,13 @@ latent_var <- kairos::latent_var %>%
   dplyr::distinct() %>% 
   dplyr::mutate(latent_var = 1)
 
+immune_pred <- kairos::immune_predictions %>% 
+  dplyr::select(specimenID) %>% 
+  dplyr::distinct() %>% 
+  dplyr::mutate(immune_pred = 1)
+
 analyses <- dplyr::full_join(jhu_tumor_file, latent_var) %>% 
+  dplyr::full_join(immune_pred) %>% 
   dplyr::mutate_at(vars(-specimenID), funs(dplyr::case_when(. == 1 ~ 1, 
                                          is.na(.) ~ 0)))