R/clustering-framework.R

In pRoloc: A unifying bioinformatics framework for spatial proteomics

setClass("ClustRegRes",
         representation(algorithm = "character",
                        parameters = "list",
                        results = "matrix",
                        criteria = "character",
                        algoparams = "list"))

setMethod("show",
          signature(object = "ClustRegRes"),
          function(object) {
            cat("Object of class \"", class(object), "\"\n", sep = "")
            cat(" Algorithm:", object@algorithm, "\n")
            cat(" Criteria:", paste(object@criteria, collapse = " "), "\n")
            cat(" Parameters:\n")
            for (i in 1:length(object@parameters)) {
              cat(" ", names(object@parameters)[i],": ")
              .param <- object@parameters[[i]]
              .n <- length(.param)
              if (.n < 7) cat(paste0(paste(.param, collapse = " "), "\n"))
              else cat(paste0(paste(.param[1:2], collapse = " "),
                              " ... ",
                              paste(.param[(.n-1):.n], collapse = " "),
                              "\n"))
            }
            invisible(NULL) 
          })

setMethod("getParams", "ClustRegRes",
          function(object, criterion) {
            if (missing(criterion)) {
              criterion <- object@criteria[1]
            } else {
              criterion <- match.arg(criterion,
                                     choices = object@criteria,
                                     several.ok = FALSE)
            }
            idx <- which.min(object@results[, criterion])
            return(object@results[idx, names(object@parameters)])
          })

setMethod("plot", c("ClustRegRes", "missing"),
          ## currently works for kmeans
          ## what if we have two parameters, like cmeans?
          ## other algorithms: spectral clustering, hyperspherical clustering
          function(x, y, ...) {
            cls <- c("steelblue", "red")
            res <- x@results[, x@criteria]            
            matplot(res, pch = 1, type = "b",
                    xlab = "k", ylab = "IC",
                    xaxt = "n",
                    col = cls)
            axis(1, at = seq_len(nrow(x@results)),
                 labels = x@parameters$k)
            grid()
            xx <- apply(res, 2, which.min)
            yy <- apply(res, 2, min)
            points(xx, yy, col = cls, pch = 19)            
            legend("topright", c("BIC", "AIC"),
                   pch = 19, col = cls, bty = "n")
            invisible(NULL)            
          })


setMethod("levelPlot", "ClustRegRes", 
          function(object, ...) {
            res <- object@results[, object@criteria]
            mins <- apply(res, 2, min)
            for (i in 1:ncol(res))
              res[, i] <- res[, i]/mins[i]
            p <- levelplot(res, xlab = "k",
                           ylab = NULL,
                           ...)
            p
          })