Curated canonical isoforms with appris

DESCRIPTION

@@ -5,7 +5,7 @@ Description: Provides precomputed data for the crisprVerse ecosystem.
  (TSSs), are provided for both mouse and human. Genomic coordinates
  of repeat elements are also provided. All objects are expressed in 
  GRCh38 (human) and GRCm38 (mouse) coordinates. 
-Version: 0.99.22
+Version: 0.99.23
 Authors@R: c(
  person("Jean-Philippe", "Fortin", email = "[email protected]", role = c("aut", "cre")),
  person("Luke", "Hoberecht", email = "[email protected]", role = c("aut"))

inst/generateTxDbData.R

@@ -71,9 +71,88 @@ canonicalMouse <- getCanonicalTranscripts(ids,
  organism="mmusculus")
 
 
+
+cleanCanonical <- function(canonicalTxs,
+ appris,
+ txObject
+){
+ cols <- c("gene_id", "tx_id", "gene_symbol")
+ tx2Gene <- mcols(txObject[["cds"]])[, cols, drop = FALSE]
+ tx2Gene <- as.data.frame(tx2Gene)
+ tx2Gene <- tx2Gene[!duplicated(tx2Gene), ]
+ rownames(tx2Gene) <- NULL
+
+ # Only keeping transcripts that exist in current annotation:
+ good <- canonicalTxs$tx_id %in% tx2Gene$tx_id
+ canonicalTxs <- canonicalTxs[good,]
+
+ # Let's add MANE select transcripts:
+ missing <- setdiff(tx2Gene$gene_id, canonicalTxs$gene_id)
+ appris <- appris[appris$gene_id %in% missing,]
+ mane <- appris[appris$mane_select,]
+ stopifnot(!any(duplicated(mane$gene_id)))
+ mane <- mane[, c("gene_id", "tx_id")]
+ canonicalTxs <- rbind(canonicalTxs, mane)
+
+ # Let's add Appris principal:
+ missing <- setdiff(tx2Gene$gene_id, canonicalTxs$gene_id)
+ appris <- appris[appris$gene_id %in% missing,]
+ principal <- appris[appris$appris_label=="PRINCIPAL",]
+ principal <- principal[order(principal$gene_id, principal$appris_number),]
+ missing <- intersect(missing, principal$gene_id)
+ wh <- match(missing, principal$gene_id)
+ principal <- principal[wh,]
+ principal <- principal[, c("gene_id", "tx_id")]
+ canonicalTxs <- rbind(canonicalTxs, principal)
+
+ # Let's add Appris alternative:
+ missing <- setdiff(tx2Gene$gene_id, canonicalTxs$gene_id)
+ appris <- appris[appris$gene_id %in% missing,]
+ missing <- intersect(missing, appris$gene_id)
+ wh <- match(missing, appris$gene_id)
+ appris <- appris[wh,]
+ appris <- appris[, c("gene_id", "tx_id")]
+ canonicalTxs <- rbind(canonicalTxs, appris)
+
+ # Are there any still missing?
+ # We will select the longest isoform for missing genes:
+ missing <- setdiff(tx2Gene$gene_id, canonicalTxs$gene_id)
+ tx2Gene <- tx2Gene[tx2Gene$gene_id %in% missing,]
+ cds <- txObject[["cds"]]
+ cds <- cds[cds$gene_id %in% missing]
+ dfs <- split(cds, f=cds$tx_id)
+ ns <- vapply(dfs, function(x) sum(BiocGenerics::width(x)), FUN.VALUE=1)
+ wh <- match(tx2Gene$tx_id, names(ns))
+ tx2Gene$len <- ns[wh]
+ dfs <- split(tx2Gene, f=tx2Gene$gene_id)
+ dfs <- lapply(dfs, function(df){
+ df <- df[order(-df$len),,drop=FALSE]
+ df[1,,drop=FALSE]
+ })
+ df <- do.call(rbind, dfs)
+ df <- df[, c("gene_id", "tx_id")]
+ canonicalTxs <- rbind(canonicalTxs, df)
+ return(canonicalTxs)
+}
+
+# Going to clean the canonical transcripts for current release:
+#load("../data/canonicalHuman.rda")
+load("../data/txdb_human.rda")
+load("../data/apprisHuman.rda")
+canonicalHuman <- cleanCanonical(canonicalHuman,
+ appris=apprisHuman,
+ txObject=txdb_human)
+#load("../data/canonicalMouse.rda")
+load("../data/txdb_mouse.rda")
+load("../data/apprisMouse.rda")
+canonicalMouse <- cleanCanonical(canonicalMouse,
+ appris=apprisMouse,
+ txObject=txdb_mouse)
+
+
+
 use_data(canonicalHuman, compress="xz", overwrite=TRUE)
 use_data(canonicalMouse, compress="xz", overwrite=TRUE)
 
 
 
-