I am using dChip for analyzing Affy Cel files. I want to find differentially expressed genes in my data. I load the cel files, normalize and log transform it. However, I am not able to get any differentially expressed genes in my samples (using 2 fold change), even though there are around 57k probe ids and the paper corresponding to the dataset reports around 5K DEGs with 5 fold change. If I do not log-transform it, I get a good number of DEGs. I am not able to understand what is happening ..kindly help!
The short answer: something is different between your methods and their methods. If you want to replicate the outcome of an analysis you need to do exactly the same as the paper you want to replicate. You have to find out, what the differences are and eliminate them.
The rest is too much guesswork for us, because there are many things that can go wrong and you neither told us what you did exactly nor which paper you are referring to. That said (and no offence), it is quite possible that you have made a simple mistake in your calculus with the logarithm.
When you want to relate fold-change cutoffs on logged values to linear fold changes you have to calculate and filter for log differences (-) not log fractions (/). For example if you want to look for genes that have two-fold change on linear scale you have to look for all genes with absolute log2-difference ≥ 1 or for an absolute log2 difference of ~2.321928, because
log(x/y) = log(x) - log(y) != log(x)/log(y)
That would explain the discrepancy.