Update GPT annotations, include those without genes

R/add_evidence.R

@@ -17,6 +17,8 @@
 #' }
 #' @param evidence_score_threshold The minimum threshold of mean
 #' evidence scores of each gene-phenotype association to keep.
+#' @param evidence_score_threshold_metric The metric to use for filtering with
+#' \code{evidence_score_threshold}.
 #' @param default_score Default evidence score to
 #' apply to gene-disease associations that are present in the HPO annotations
 #' but don't have evidence scores in the GenCC annotations.
@@ -37,16 +39,17 @@
 #' phenos2 <- add_evidence(phenos = phenos)
 add_evidence <- function(phenos,
                          evidence_score_threshold = NULL,
+                         evidence_score_threshold_metric="evidence_score_sum",
                          all.x = TRUE,
                          allow.cartesian = FALSE,
                          agg_by = c("disease_id",
                                     "gene_symbol"),
                          default_score = 1,
                          ...){
-  evidence_score <- evidence_score_mean <- NULL;
+  evidence_score <- NULL;
 
   if(!all(c("evidence_score_mean") %in% names(phenos))){
-    messager("Annotating gene-disease associations with Evidence score")
+    messager("Annotating gene-disease associations with Evidence Score")
     phenos <- add_disease(phenos = phenos,
                           all.x = all.x,
                           allow.cartesian = allow.cartesian)
@@ -63,11 +66,13 @@ add_evidence <- function(phenos,
   }
   #### Filter ####
   if(!is.null(evidence_score_threshold)){
-    if("evidence_score_mean" %in% names(phenos)){
-      phenos <- phenos[evidence_score_mean>=evidence_score_threshold,]
-    } else if("evidence_score" %in% names(phenos)){
-      phenos <- phenos[evidence_score>=evidence_score_threshold,]
+    if(!evidence_score_threshold_metric %in% names(phenos) &&
+       "evidence_score" %in% names(phenos)){
+      messager(evidence_score_threshold_metric,"not found in phenos.",
+               "Using evidence_score for filtering step instead.")
+      evidence_score_threshold_metric <- "evidence_score"
     }
+    phenos <- phenos[get(evidence_score_threshold_metric)>=evidence_score_threshold,]
   }
   #### Set default score ####
   if(!is.null(default_score)){

R/add_hpo_id.R

@@ -2,16 +2,18 @@
 #' Add HPO ID column to dataframe
 #'
 #' Adds the HPO term ID column "hpo_id".
+#' @inheritParams KGExplorer::map_ontology_terms
 #' @export
 #' @examples
 #' phenotype_to_genes <- load_phenotype_to_genes()
 #' phenos <- unique(phenotype_to_genes[,c("hpo_id","hpo_name")])
 #' phenos2 <- add_hpo_id(phenos=phenos)
 add_hpo_id <- function(phenos,
-                       hpo = get_hpo()) {
+                       hpo = get_hpo(),
+                       ignore_case = TRUE) {
   if(!"hpo_id" %in% names(phenos)){
-    messager("Adding HPO IDs.")
     phenos$hpo_id <- map_phenotypes(hpo = hpo,
+                                    ignore_case = ignore_case,
                                     terms = phenos$hpo_name,
                                     to = "id")
   }

R/get_hpo.R

@@ -9,7 +9,7 @@
 #' Note that the maximum ontology level depth in the 2016 version was 14,
 #' whereas in the 2023 version the maximum ontology level depth is 16
 #'  (due to an expansion of the HPO).
-#' @inheritParams KGExplorer::add_ancestors
+#' @inheritParams KGExplorer::get_ontology
 #' @inheritDotParams KGExplorer::get_ontology
 #' @returns \link[simona]{ontology_DAG} object.
 #'

R/gpt_annot_read.R

@@ -2,38 +2,46 @@
 #'
 #' Read in phenotype annotations generated by GPT and
 #'  do some initial preprocessing (e.g. adding HPO IDs).
-#' @param path Path to annotations CSV file.
-#'  If \code{NULL}, will pull data from GitHub Releases instead.
+#' @inheritParams main
+#' @param save_path Path to annotations CSV file.
+#'  If the file does not exist, the data will be downloaded from GitHub.
+#' @param force_new If \code{TRUE}, the data will be downloaded from GitHub
+#' even if it already exists locally.
 #' @param verbose Print messages.
-#' @source code {
-#' piggyback::pb_upload(file = "~/Downloads/gpt4_hpo_annotations.csv",
-#'                      repo = "neurogenomics/HPOExplorer",
-#'                      overwrite = TRUE,
-#'                      tag = "latest")
-#' }
+#' @param include_nogenes Include phenotypes with no associated genes.
 #' @returns data.table of phenotype annotations
 #'
 #' @export
 #' @examples
 #' annot <- gpt_annot_read()
-gpt_annot_read <- function(path = NULL,
+gpt_annot_read <- function(save_path=file.path(
+  KGExplorer::cache_dir(package = "HPOExplorer"),
+  "gpt4_hpo_annotations.csv"
+),
+                           force_new=FALSE,
+hpo=get_hpo(),
+include_nogenes=TRUE,
                            verbose=TRUE){
-  pheno_count <- hpo_name <- hpo_id <- NULL;
+  pheno_count <- hpo_name <- hpo_id <- phenotype <- NULL;
 
-  if(is.null(path)){
-    # path <- paste0(
-    #   "https://github.com/neurogenomics/RareDiseasePrioritisation/raw/master/",
-    #   "gpt_annotations/gpt4_hpo_annotations.csv"
-    # )
-    path <- get_data("gpt4_hpo_annotations.csv")
+  if(!file.exists(save_path) || isTRUE(force_new)){
+    path <- paste0(
+      "https://github.com/neurogenomics/RareDiseasePrioritisation/raw/master/",
+      "gpt_annotations/gpt4_hpo_annotations.csv"
+    )
+    utils::download.file(path, save_path)
+    # path <- get_data("gpt4_hpo_annotations.csv")
   }
-  d <- data.table::fread(path, header = TRUE)
+  d <- data.table::fread(save_path, header = TRUE)
+  d <- d[!is.na(phenotype)]
   data.table::setnames(d,"phenotype","hpo_name")
+  d <- add_hpo_id(d, hpo = hpo)
   #### Check phenotype names ####
   annot <- load_phenotype_to_genes(verbose = verbose)
   d <- merge(d,
              unique(annot[,c("hpo_id","hpo_name")]),
-             by="hpo_name")
+             all.x = TRUE,
+             by=c("hpo_name","hpo_id"))
   d <- data.frame(d)
   d[d==""] <- NA
   d <- data.table::data.table(d)

R/map_phenotypes.R

@@ -1,14 +1,17 @@
-#' @describeIn map_ map_
 #' Harmonise phenotypes
 #'
 #' Harmonise a mixed vector of phenotype names (e.g. "Focal motor seizure")
 #' and HPO IDs (e.g. c("HP:0000002","HP:0000003")).
+#' @inheritParams map_
+#' @inheritParams KGExplorer::map_ontology_terms
 #' @returns Character vector
 #'
 #' @export
 #' @import KGExplorer
 #' @examples
-#' terms <- c("Focal motor seizure","HP:0000002","HP:0000003")
+#' terms <- c("Focal motor seizure",
+#'             "Focal MotoR SEIzure",
+#'             "HP:0000002","HP:0000003")
 #' #### As phenotype names ####
 #' term_names <- map_phenotypes(terms=terms)
 #' #### As HPO IDs ####
@@ -17,10 +20,12 @@ map_phenotypes <- function(terms,
                            hpo = get_hpo(),
                            to=c("name","id"),
                            keep_order = TRUE,
+                           ignore_case = TRUE,
                            invert = FALSE){
   KGExplorer::map_ontology_terms(terms = terms,
                                  ont = hpo,
                                  to = to,
                                  keep_order = keep_order,
+                                 ignore_case = ignore_case,
                                  invert = invert)
 }

R/plot_evidence.R

@@ -81,8 +81,8 @@ plot_evidence <- function(metric="evidence_score_sum",
   pw <- patchwork::wrap_plots(h1,h2,h3,h4,h5, ncol = 1)
   if(isTRUE(show_plot)) methods::show(pw)
 
-  KGExplorer::plot_save(plt = pw,
-                        path = save_path,
+  KGExplorer::plot_save(plt=pw,
+                        save_path=save_path,
                         height=height,
                         width=width)
   return(