I have been analysing a set of arrayCGH data from tumour samples. My output is a matrix of genomic regions and their call states (-1,0,1,2; representing losses, no-change, gains and amplifications in those regions) for each sample.
To identify significant regions between two sample groups I have used CGHtest, an R package that has allowed me to run 10K permutations and generate a corrected p-value using Kruskal-Wallis / Willcoxon / chi-squared testing.
However I would like to see what other phenotypes are influencing the significant regions, or rather eliminate the effect of tumour aggressiveness measures and age of patient etc etc. There is a package CNVassoc that can covary in the way that I would like but I have been unsuccessful using it for my data (and for the example data) and also it doesn't perform permutations.
Can anyone recommend a method or package for doing what I would like? I am rather new to R and stats is not really my territory so please excuse me if there are really obvious answers to this question.
