Dear all, 

please could you advise on a package/ functions that do more RNA-seq quality controls, or check the distribution of the data. I am encountering the following situation with RNA-seq analysis :

A-- i have 2 sets of replicates, these are labelled : siCTL (replicate1,replicate2) vs siPROTEIN (replicate1, replicate2).the correlation coefficient between replicates is 0.99 .

B-- shall I take any replicate individually, and just compute the fold-changes, there are approx 1000 up-reg genes, and 1000 down-reg genes 

C-- however, considering both replicates, edgeR or limma give only approx 200 up-reg and 200 down-reg genes (fdr<0.05). even a simple t-test will give similar results.  

how shall i understand the difference between the situation B and situation C ? RNA quality and data statistical distribution ? other factors ? thank you, 

-- bogdan

re A: A correlation coefficient of 0.99 is really quite high. Is this really biological replicates? Is the data log transformed? Still, a high correlation coefficient has little meaning as a quality control. Even between different samples you expect high correlation, because regardless of experimental condition, genes tend to stay in a certain expression range (transcription factors will always be rather low, genes for structural proteins like actin or for ribosomal proteins will always be rather high). Still, you want to see more correlation between replicates than between differently treated samples, of course.

re B: I am surprised that there are any genes that are neither up-regulated nor down-regulated. You really have genes that do not change at all, i.e., have exactly the same expression in the compared samples? Or have you used a threshold? (If so: When asking questions on this forum,. please take care to describe comprehensively and accurately what you have done.)