R/predict_most_similar_cluster.R

In JCVenterInstitute/FRmatch: Cell type matching in single-cell RNA-sequencing data using FR-Match

#' Prediction function for FR-Match results
#'
#' This function takes in the \code{\link[FRmatch]{FRmatch}} output and predicts the most similar reference cluster for each query cluster.
#'
#' @param rst.FRmatch The \code{\link[FRmatch]{FRmatch}} output.
#' @param p.adj.method P-value adjustment method for multiple hypothesis testing correction. Default: \code{"BY"}.
#' For more options, please see \code{\link[stats]{p.adjust.methods}}.
#'
#' @return A data frame of the most similar reference cluster for each query cluster,
#' which is the "match" with the highest adjusted p-value. If there are tied p-values, all will be returned.
#'
#' @export

predict_most_similar_cluster <- function(rst.FRmatch, p.adj.method="BY"){
  pmat <- rst.FRmatch$pmat
  pmat.adj <- padj.FRmatch(pmat, p.adj.method=p.adj.method)

  ## return ALL maximum p-values if there are multiple
  cmax <- colMaxs(pmat.adj)
  query_cluster <- most_similar_ref_cluster <- padj <- c()
  for(j in 1:length(cmax)){
    ind.j <- pmat.adj[,j] == cmax[j]
    query_cluster <- c(query_cluster, rep(colnames(pmat.adj)[j], sum(ind.j)))
    most_similar_ref_cluster <- c(most_similar_ref_cluster, rownames(pmat.adj)[ind.j])
    padj <- c(padj, rep(cmax[j], sum(ind.j)))
  }
  df.most.similar.cluster <- data.frame(query_cluster, most_similar_ref_cluster, padj) %>% arrange(desc(padj))

  # ## return the first maximum p-value
  # df.most.similar.cluster <- data.frame("query_cluster"=colnames(pmat),
  #                                       "most_similar_ref_cluster"=rownames(pmat)[apply(pmat, 2, which.max)], #use the original pmat, independent of p.adj.method
  #                                       "padj"=colMaxs(pmat.adj)) %>% arrange(desc(padj)) #return padj here

  return(df.most.similar.cluster)
}