log transformation in the analysis of gene expression by chip-array

Question

[arielbier]

Hi everyone I have a question regarding to the analysis of gene expression by chip-array. When I am analyzing the ratio of two groups of samples by making average of all the samples in the each group, do I need to do log transformation for each sample and then to calculate average and the ratio or to calculate average and the ratio and then to do log transformation. For example, if I have in group1 the values: a,b and c and in group 2: d, e and f, do I need to do- log2[average(a,b,c)/average(d,e,f)] OR average(log2 a, log2 b, log2 c)/ average(log2 d, log2 e, log2 f)? If someone can help me with this question I will be very grateful.

Answer 1

[Kevin Blighe]

Better late than never:

If you have processed your data by the Robust MultiArray Average (RMA) method and your array is single-colour, then your data is already log [base 2] (log2) transformed. Proceed with analyses via Limma, which will derive log2 ratios for you, coupled with p-values.

If your array is 2-colour, then your data is already representative of log2 ratios between colour 1 vs. colour 2. The analysis approach in this case may differ; however, it's more than likely that you have a single colour array.

Note that the calculation is this:

a=c(1,2,3,4,5)
b=c(4,5,6,7,8)

log2(mean(b) / mean(a))
[1] 1

...or this:

log2(mean(b)) - log2(mean(a))
[1] 1

Kevin