cuffdiff "value_1" & "value_2", are they mean FPKM values from replicates for each condition?

Question

[vivekananthrp]

Hi, I have run cuffdiff for finding differentially expressed genes in two conditions. Each condition has 3 biological replicates, which were processed using cufflink. The "value_1" and "value_2" columns in gene_exp.diff file don't reflect the FPKM values from cufflinks. The FPKM values from 3 replicates are, 21.168, 39.8954 and 17.9224. But the value_1 for the gene from gene_exp.diff file is 30.3461 (which is definitely not the average of three values) . How does cuffdiff generates 'value_1' and 'value_2' ?

I am using latest versions of cufflink and cuffdiff.

Thanks in advance

Answer 1

[Devon Ryan]

The cuff* programs are a black box, no one other than the authors really knows what those programs are doing internally (sure, you can go through the code, but are you really going to do that for C++ code?).

Anyway, there's at least supposed to be some normalization that takes place inside of cuffdiff before it does its comparisons. value_1 and value_2 presumably come from those normalized values.