Load the gene data and transpose it

# Load the gene dataset
gene <- read.delim("~/Google Drive/CR Rao Course/bodygene.csv")
gene[c(1:2),c(1:7)]
##        Gene  adipose    adrenal       brain   breast     colon heart
## 1 1060P11.3 0.762865 0.00000000 0.000000000 0.971663 0.0000000     0
## 2      A1CF 0.000000 0.02799675 0.000288957 0.000000 0.3955925     0
# Transpose the dataset
gene <- t(gene)
# Convert the matrix to a dataframe. 
gene <- as.data.frame(gene)

Nearest neighbour single linkage cluster analysis

Calculate the eucledian distance between the observations.

gene_dist <- dist(gene [-1, ])

Hierarchical Clustering

gene_clust <- hclust(gene_dist)

Dendrogram

plot(gene_clust, hang = -1)

K-means clustering

We use the k-means command to generate the clusters

gene_clust2 <- kmeans(gene[-1, ], centers = 3, nstart=25)

Evaluating the cluster generated

gene_clust2$cluster
##        adipose        adrenal          brain         breast          colon 
##              1              1              1              1              1 
##          heart         kidney          liver           lung      lymphnode 
##              1              1              2              1              1 
##          ovary       prostate skeletalmuscle         testes        thyroid 
##              1              1              3              1              1 
##     whiteblood 
##              1

Create a loop to determine optimal K

# Creata a folder with 10 0s.
ess <- rep(0,10)
#Create the loop for this command
for (k in 1:10) 
{
  ess[k] <- kmeans(gene[-1, ],nstart = 25,centers=k+1)$tot.withinss # in this we say that the k value of ess is the kmeans cultering algorirthm for K+1 centers and get only the total ESS value.
}

Determining the optimal number of K using scree plot

plot(2:11, ess, xlab = "Number of clusters",ylab="Within Sum of Squares", col="blue", pch=10, main ="Scree Plot to determine optimal K")

The optimal value of K seems to be 5 or 6
We do k-means clustering with 5 clusters

gene_clust3 <- kmeans(gene[-1, ], centers = 5, nstart=25)

Check the cluster

gene_clust3$cluster
##        adipose        adrenal          brain         breast          colon 
##              3              5              1              3              5 
##          heart         kidney          liver           lung      lymphnode 
##              1              1              2              5              5 
##          ovary       prostate skeletalmuscle         testes        thyroid 
##              1              1              4              1              1 
##     whiteblood 
##              5