[BioC] Clustering non-biological replicate chips

[Ken Termiso]

Hi all,

I have an experiment with several control and several experimental replicate 
chips, and the chips are not actual biological replicates, but rather 
represent replicate hybridizations or amplifications. In terms of the data 
to be input into a clustering algorithm such as kmeans, would it be better 
for me to calculate a single average value for the control side and one for 
the experimental side than it would be to input all of the replicate chip 
data? It seems that using all of the replicate data points from 
non-biological replicates would only help partition non-biological aspects 
of the data and not help answer questions regarding what is happening 
biologically between the control and experimental cases.

Thanks in advance,
Ken