# This function does simple clustering in the sample direction
# for an exprSet. It alwyas works with log-transformed data. The
# heuristic for determining if the data has been log-transformed
# is crude: is just checks to see if any value is bigger than 100.
# Expression filtering is automatic; we only keep the genes whose
# mean expression is greater than the 75th percentile. We also
# give no options for distance metric or linkage;  all clustering
# is performed with Pearson correlation distance and average linkage.

simpleCluster <- function(es) {
  if (!inherits(es, 'exprSet')) {
    stop('This function only works for exprSet objects')
  }
  data <- exprs(es)
  if (max(data, na.rm=TRUE) > 100) { # needs to be log transformed
    data <- logb(data, 2)
  }
  avg <- apply(data, 1, mean) # get mean expression of each gene
  data <- data[avg > quantile(avg, 0.75), ] # only keep the top 25%
  distmat <-  as.dist((1-cor(data))/2) # distance based on correlation
  plclust(hclust(distmat, "average"))
  invisible(0)
}