predicting replicates from a pooled sample - (Jun/12/2009 )

Hi, I have run 2 affy chips for pooled samples. the samples initially had 6 replicates which were pooled into one sample for each chip. This was initally to keep cost down.

I was wondering if anyone knew of a statistical (or any other way) method to predict reproducibility and predict measures of how data may be spread if the samples had been run on 12 chips (6 for each group)?

Is there anyway that I can verify reproducibility without running extra chips?

Greenstripe

I am pretty certain that this is impossible; you are trying to find out the reproducibility of an assay having run only one sample (that you pooled several samples together before does not really matter in that respect).

You can make broad assumptions about the individual samples assuming they were perfect replicates judging from your noise distribution but that's about it.