How can I (preferably in R/Bioconductor) perform Hierarchical clustering on CBS-segmented tumor-samples (CNV analysis)?
I have copy number variation data on 50 tumor samples. I have already preprocessed them to Log R ratios and CBS-segmented them. For each tumor sample, I have information available on variables like "tumor stage", "hasMutationInGeneX", "isPrimaryTumor", "ageOfDiagnosis" etcetera. I am interested if I can find groups of tumors (patients) that share similar copy number profiles so I want to do hierarchical clustering on the CBS-segmentation output.
Now I know their are lots of hierarcical clustering programs for microarray-expression data but I think I cannot use them. The point is that for microarray expression the output can be stored in a matrix where each row represents a shared feature by all samples:
probes Sample1 Sample2 SampleN
10101_at 0.2 0.7 -0.2
01010_at -0.3 -0.2 -1.2
However, concerning CNV-analysis the segmentation on different samples does not have features like probes that are measured in each sample, in stead you get something like this:
chr start end sample meanCopyNumber
1 100 8000 Sample1 0.2
1 200 500 Sample2 0.3
Though I am aware of packages like reshape can slice and dice data frames to get them in the desired format, I wonder if there are Biocondcutor packages that take CBS-output and do hierarchical clustering (and maybe even heatmap + dendrogram visualization??)
1 answer
Instead, try creating a matrix of genes vs samples. For each value, print the copy number of that gene in that sample. Alternately, use genomic windows of, say 10k and assign each window the average copy number of the probes within it.
Log in to answer this question.