# [R] Visualizing and clustering one half of a symmetric matrix

Khan, Saad M. (MU-Student) smk5g5 at mail.missouri.edu
Fri Sep 16 06:40:18 CEST 2016

```I do want to cluster it and only plot the lower half of the matrix.

________________________________
From: Peter Langfelder <peter.langfelder at gmail.com>
Sent: Thursday, September 15, 2016 11:33:13 PM
Cc: r-help at R-project.org
Subject: Re: [R] Visualizing and clustering one half of a symmetric matrix

Do not set the upper (or lower) triangle to NA. Simply supply the full
matrix to pheatmap. I am not an expert on pheatmap but looking at the
manual you should supply clustering_distance_rows = "none",
clustering_distance_cols = "none" or something like that to make
pheatmap interpret the matrix as a distance matrix. Read carefully
through the help on pheatmap to make sure the function plots what you
want it to plot.

HTH,

Peter

On Thu, Sep 15, 2016 at 7:38 PM, Khan, Saad M. (MU-Student)
<smk5g5 at mail.missouri.edu> wrote:
> Hi all,
>
> I have a distance matrix (symmetric) which looks somewhat like this (only a small portion shown)
>
>                 ENSG00000101413 ENSG00000176884 ENSG00000185532 ENSG00000106829
> ENSG00000101413           1.000           1.000           1.000           1.000
> ENSG00000176884           0.328           0.258           0.260           0.390
> ENSG00000185532           1.000           1.000           1.000           1.000
> ENSG00000106829           0.684           0.443           0.531           0.701
>
> These distances are custom measures that I need to cluster. Since it's a symmetric matrix I only need to consider one half triangle of the matrix. So I do something like this :-
>
> newmat <- ensembl_copygosimmat
> newmat[upper.tri(ensembl_copygosimmat)] <- NA
>
> Then I wanted to visualize how the lower triangle looked using pheatmap which does hierarchical clustering itself.
>
> library(pheatmap)
> pheatmap(newmat)
>
> But since there are NA values in the matrix (in the upper half) it always throws an error. I was wondering what would be the ideal way to visualize as well as cluster such a matrix.
>
> Regards
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help