Hi board,

We are currently interested in using limma to identify whether the ratios between the expression of two genes are different between two groups of individuals. Is it sound to generate an extra row of Exp(A)/Exp(B) on the expression matrix and progress with the usual limma pathway, or should I do it differently?

Thanks for your advice in advance!

Cheers,
John

Yes, if you have a matrix of normalized log2-expression values, you could add a new row equal to the difference between gene A and gene B. The new row would contain log2( exp(A) / exp(B) ).

It is better to work from the normalized log-expression values. Please don't try to manipulate unlogged and unnormalized expression values.

I am assuming you have microarray data.

If in fact you have RNA-seq data, then the approach needs more thought. In that case you could just use logCPM(A) - logCPM(B) as the dependent variable, but it would need to be analysed separately as a univariate statistical analysis rather than added as an extra row to other expression data.