BME 210 Lab 7

Part II

hyper.overlap <- function(gold.matrix, cluster.assign)
{
   clusters <- unique(cluster.assign)
   num.clusters <- length(clusters)
   num.gold <- dim(gold.matrix)[2]
   pvals <- matrix(nrow=num.gold, ncol=num.clusters, data=0)

   for(g in 1:num.gold)
   {
      gold.genes <- which(gold.matrix[,g]!=0)
      for(c in 1:num.clusters)
      {
         cluster.genes <- which(cluster.assign == c)
         overlap <- length(intersect(gold.genes, cluster.genes))
         num.genes <- length(cluster.assign)
         white.balls <- length(gold.genes)
         black.balls <- num.genes - white.balls
         k <- length(cluster.genes)
         pvals[g,c] = phyper(q=overlap-1, m=white.balls, n=black.balls, k=k,
                lower.tail=FALSE, log.p = TRUE) / log(10)
      }
   }

   return(pvals)
}

There isn't much of a one-to-one correspondence between the K-Means clusters and KEGG categories. Many KEGG categories that hit

Part III