CODEWITHSUNDEEP
cluster-analysis
Nicolas Rosewick
Nicolas Rosewick

Reputation: 1998

Clustering using two matrix

I've two matrix containing informations from 40 samples and 50000 genes. Matrix Expr contains the gene expression for each gene and samples; Matrix Methyl contains the methylation state of these genes for each samples. Is it possible to perform a clustering (on genes and/or samples) based on both expression and methylation informations ? I know how to perform a basic hierarchical clustering in R i.e; hclust(dist(M)) but it's only on one matrix.. Any idea/advice ?

Upvotes: 0

Views: 344

Answers (2)

Has QUIT--Anony-Mousse
Has QUIT--Anony-Mousse

Reputation: 77454

You need to define a similarity that takes both matrixes into account.

Naively, this could be as easy as

dist <- dist(A) + dist(B)

however, clustering in general is extremely sensitive to scale, and these problem make any such approach very difficult. Sorry - there is no "correct" or automatic solution to this problem.

Upvotes: 1

Victor Pecanins
Victor Pecanins

Reputation: 11

If you want to cluster Samples based on their (dis)similarities taking into account gene expression and methylation state, then you can consider that Gene Expression and Gene Methylation State for all the 50000 genes are all 'Features' of each sample.

So, you can concatenate both matrices Methyl and Expr, resulting in a 40x100000 matrix, and computing the dist() of that matrix.

Similarly, in case you want to cluster genes based on their differences, you can concatenate both matrices onto a 80x50000 matrix

Hope it helps.

Upvotes: 0

Related Questions