An HTML attachment was scrubbed... URL: https://www.stat.math.ethz.ch/pipermail/bioconductor/attachments/ 20040720/34a4be92/attachment.html

Arne I don't have the exact code to hand, but as I recall heatmap() either takes data and does the clustering itself, or it takes two pre- computed trees, as from the output of hclust(). Therefore, you need to do two hclust() operations to produce two trees, and then pass those trees as arguments to heatmap(). Mick -----Original Message----- From: Arne.Muller@aventis.com [mailto:Arne.Muller@aventis.com] Sent: 20 July 2004 15:28 To: bioconductor@stat.math.ethz.ch Subject: [BioC] heatmaps and pearson correlation Hello, I'm trying to construct a heatmap by clustering two dimensions (genes, treatments) using pearson correlation for both the distances (1 - correlation). Unfortinately the hclust and dist methods don't provide the pearson as a metric for distances (is there any reason for this?). Has anyone a hint on how to make hclust and heatmap to cluster both dimensions via pearson correlation? thanks a lot for help, Arne [[alternative HTML version deleted]]

Hi! If your data are in the matrix use cor. cor returns you the correlation matrix. use ?t (transpose) to compute the simmilarities either for rows or for columns. mydist<-as.dist(1-cor(mymatrix)) now you can pass it to hclust. Hope it helps. *********** REPLY SEPARATOR *********** On 7/20/2004 at 4:27 PM Arne.Muller@aventis.com wrote: Hello, I'm trying to construct a heatmap by clustering two dimensions (genes, treatments) using pearson correlation for both the distances (1 - correlation). Unfortinately the hclust and dist methods don't provide the pearson as a metric for distances (is there any reason for this?). Has anyone a hint on how to make hclust and heatmap to cluster both dimensions via pearson correlation? thanks a lot for help, Arne Dipl. bio-chem. Eryk Witold Wolski @ MPI-Moleculare Genetic Ihnestrasse 63-73 14195 Berlin 'v' tel: 0049-30-83875219 / \ mail: wolski@molgen.mpg.de ---W-W---- http://www.molgen.mpg.de/~wolski [[alternative HTML version deleted]]

Thanks for your help, this is exactly what I was looking for. However, now I've a problem with the colors of the heatmap. Tme values in the heatmap matrix range from 0 to 17 were 1 is neutral. I'd like to have a gradient from blue (values 0 to >1) to grey (neutral, here 1) and red (value >1 to 17). I've tried maPalette from the marray package, the colors don't seem to be assigned as I'd like them (the value range from 0 to 17 of my data matrix gets distributed evenly overthe color range). I'm happy for any suggestions, on how to construct the gradients, maybe there's already a package available ... thanks again for your help, Arne -----Original Message----- From: Wolski [mailto:wolski@molgen.mpg.de] Sent: 20 July 2004 16:45 To: Muller, Arne PH/FR; bioconductor@stat.math.ethz.ch Subject: Re: [BioC] heatmaps and pearson correlation Hi! If your data are in the matrix use cor. cor returns you the correlation matrix. use ?t (transpose) to compute the simmilarities either for rows or for columns. mydist<-as.dist(1-cor(mymatrix)) now you can pass it to hclust. Hope it helps. *********** REPLY SEPARATOR *********** On 7/20/2004 at 4:27 PM Arne.Muller@aventis.com wrote: Hello, I'm trying to construct a heatmap by clustering two dimensions (genes, treatments) using pearson correlation for both the distances (1 - correlation). Unfortinately the hclust and dist methods don't provide the pearson as a metric for distances (is there any reason for this?). Has anyone a hint on how to make hclust and heatmap to cluster both dimensions via pearson correlation? thanks a lot for help, Arne Dipl. bio-chem. Eryk Witold Wolski @ MPI-Moleculare Genetic Ihnestrasse 63-73 14195 Berlin 'v' tel: 0049-30-83875219 / \ mail: wolski@molgen.mpg.de ---W-W---- http://www.molgen.mpg.de/~wolski