Hierarchical Clustering of copy number variations

Question

[kulani.s91]

I have a dataset of CNA of genes

GENE X1 X2 X3 X4 X5
TP53 0  1  0  0  -2
PTEN 0  0 -1  0  0
AR   NA NA NA NA NA

When I run the code,

> mRNA_Data_hclust <- dist(test, method='euclidean') 
> mRNA_hclust <- hclust(mRNA_Data_hclust, method = "complete")

I get the following error.

Error in hclust(mRNA_Data_hclust, method = "complete") : 
  NA/NaN/Inf in foreign function call (arg 11)

How do I handle the error without removing the NA values?

[Sean Davis]

Why not just remove NA values?

[Jan Oosting]

CNA from RNA data has been done in the past, but is very challenging

You could use impute() to cover up NAs. but that will not help with entire rows of NAs

[Kevin Blighe]

Important to understand why something is NA, in this case. You ave other values that are 0, which indicates unchanged, i.e., normal copy number. What's the difference between NA and 0 in your data?

I think that the ideal solution is to investigate what leads to a NA being assigned.