I asked related question previously @ log ratio in gene expression array
The question I have is when we calculate logA- logB it is different from logA/logB assuming A and B are two groups of genes, the FC is different. Will not logFC or linear FC same in both cases. Am I missing some importrant point. This appear to be a naive question but I get stuck.
Log ratios should be presented as signed values, where negative log-ratios indicate down-regulation and positive ones represent up-regulation. If your pre-transformed values are A and B you can compute the log-ratio as log(B)-log(A) or log(B/A), both are equivalent, assuming of course your values A and B are positive. Referring to your sentence in your question - you would not expect log(A)-log(B) to be the same as log(A)/log(B), those are two unrelated calculations. If you have replicate gene expression values that you are averaging to compute a log fold change, you should log transform the individual values prior to averaging.
