R/makeQueries.R

In netDx: Network-based patient classifier

#' Randomly select patients for queries for feature selection
#' 
#' @param incPat (char) vector of patient IDs to be included in query
#' @param featScoreMax (integer) Number of times to run query, usually equal 
#' to the max score for features in the design (e.g. if featScoreMax=10, then 
#' this value is 10).
#' @param verbose (logical) print messages
#' @return (list) of length \code{featScoreMax}, containing names of patients 
#' in query file for each fold
#' @examples 
#' data(pheno)
#' x <- makeQueries(pheno$ID)
#' @export
makeQueries <- function(incPat, featScoreMax = 10L, verbose = TRUE) {
    
    # randomly reorder for N-fold partitioning.
    incPat <- sample(incPat, replace = FALSE)
    # num in query file
    num2samp <- floor(((featScoreMax - 1)/featScoreMax) * length(incPat))
    # num to retrieve from GM database in each iteration
    csize <- round((1/featScoreMax) * length(incPat))
    
    if (verbose) {
        message(sprintf("\t\t%i IDs; %i queries (%i sampled, %i test)", 
						length(incPat), 
            featScoreMax, num2samp, csize))
    }
    
    out <- list()
    for (k in seq_len(featScoreMax)) {
        sidx <- ((k - 1) * csize) + 1
        eidx <- k * csize
        if (k == featScoreMax) 
            eidx <- length(incPat)
        p1 <- sprintf("\t\tQ%i: %i test; ", k, eidx - sidx + 1)
        
        out[[k]] <- setdiff(incPat, incPat[sidx:eidx])
        if (verbose) 
            message(sprintf("%s %i query", p1, length(out[[k]])))
    }
    
    out
}