Clustering differences between heatmap.2 and pheatmap
Entering edit mode
6.4 years ago
igor 12k

I have been using heatmap.2 for a while, but just discovered pheatmap. In heatmap.2, you can specify clustering settings via distfun and hclustfun. In pheatmap, you have clustering_distance_rows and clustering_method. However, if I set those parameters to use the same algorithms, the resulting heatmaps do not look similar. How can that be? Does pheatmap perform additional manipulations that heatmap.2 does not?

My code:

# pheatmap
pheatmap(vals, color=colors, scale="row", cluster_rows=T, cluster_cols=T, clustering_distance_rows = "euclidean", clustering_distance_cols = "euclidean", clustering_method = "complete")

# heatmap.2
hclust_fun = function(x) hclust(x, method="complete")
dist_fun = function(x) dist(x, method="euclidean")
heatmap.2( as.matrix(vals), scale="row", trace="none", dendrogram="both", Rowv=TRUE, Colv=TRUE, distfun=dist_fun, hclustfun=hclust_fun, col=colors)
heatmap R • 13k views
Entering edit mode
5.9 years ago
Lerong ▴ 110

Basically when you show scaled data, heatmap.2 scale data after clustering , whereas pheatmap scales data before clustering. I am guessing that makes the difference in the final output sometimes.

Entering edit mode

A related thread about scaling data and these 2 heatmap functions: cannot replicate the pheatmap scale function

Entering edit mode

Sorry to comment on this old post, I also just have noticed this difference when trying to create heatmaps. Do you have any suggestion on which is the better way? Clustering then scaling (like heatmap / heatmap.2) or scaling then clustering (like pheatmap), because the cluster results is quite different.

Entering edit mode
6.1 years ago

heatmap.2 applies some reordering to the dendrogram that is not done by pheatmap. Here is an excerpt from heatmap.2 manual:

If either is a vector (of "weights") then the appropriate dendrogram is reordered according to the supplied values subject to the constraints imposed by the dendrogram, by reorder(dd, Rowv), in the row case. If either is missing, as by default, then the ordering of the corresponding dendrogram is by the mean value of the rows/columns, i.e., in the case of rows, Rowv <- rowMeans(x, na.rm=na.rm). If either is NULL, no reordering will be done for the corresponding side.

I decided to specify the clustering method for both rows and columns in heatmap.2

distance.row = dist(as.matrix(vals), method = "euclidean")
cluster.row = hclust(distance.row, method = "ward.D")
distance.col = dist(t(as.matrix(vals)), method = "euclidean")
cluster.col = hclust(distance.col, method = "ward.D")
heatmap.2(vals, scale="row",trace="none", dendrogram="both", Rowv=as.dendrogram(cluster.row), Colv=as.dendrogram(cluster.col))

The order now is more similar to pheatmap, but not completely identical...


Login before adding your answer.

Traffic: 2152 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6