How to find DE genes using RPKM values?

[casshyr]

Hello,
I have essentially 2 datasets, each showing the RPKM values for each gene. I want to compare the gene expressions from one set to another, and see which ones are up-regulated or down-regulated significantly comparing to the other.
I have tried some primitive ways, such as dividing the two by each other and see if that ratio is greater than a fold change threshold..but this yields to me like 10000 genes, which is unlikely.

Are there any suggestions on how to find differentially expressed genes based on RPKM values?

btw I don't have access to the mapped reads data, so programs like DEGseq won't work for me. I only have access to the RPKM values

Thanks!

[Simon Anders]

Well, if you get to many genes, you have to make your fold-change threshold stricter. As this is all you have, you cannot do better.

What would be a good threshold? Obviously one which reflects how much the RPKM value for a gene typically change between two samples from the same experimental condition. Only if you fold change is much stronger, you can assume that the change is due to the change in condition.

As you don't have replicates, you can only guess what a good threshold is. In other words: No, you cannot get any reasonable results from just two samples.

Simon

[mgogol]

Maybe you could use regression and outlier detection to find outliers in the plot of one data set vs. another (the points furthest from the regression line are the genes showing the biggest changes).

You should probably look at some histograms and set a lower cutoff on the RPKMs you're willing to consider if you use fold change, because using fold change with really small RPKM values will be very noisy.

Once you have a ranked list of genes by RPKM fold-change, you could test using qPCR to verify your results.