RNA seq bioinformatics

[renuka22]

Hello all,

I have RNA seq data from 2 replicates of control and treated (say for example C1,C2,T1,T2). But when I compared the replicates separately (C1vsT1) ,(C2vsT2) there were around 1500 differentially expressed genes, but when the company combined the replicates together (all control C1C2 together vs T1T2 together) the differentially expressed genes were around 160 only. Also, the differentially expressed genes which showed up for combined replicates were different than the ones in C1TI or C2T2. Could anyone tell me what might be the reason for this?

Thanks

[Biocomputronics]

It all depends on how you are doing the test for differential gene expression. But the basic idea is around statistical power and significance. Generally speaking, you cannot compare one replicate for treatment A with one replicate for treatment B since you have zero statistical power. Even 2 replicates is generally considered not enough since there is a lot of biological and technical variation involved. Most people suggest at least 3 replicates per condition. Then you need to use a principled statistical method to determine differential expression such as the one's incorporated in the packages edgeR or DESeq2. You can't really use simplistic tests like a t-test with RNA-seq data.