DESeq2-issues with input data

Question

[vivek37373]

I have a small rna dataset of 12 samples containing two replicates for specific conditions. For example,

CON1 ---> 30 million reads
CON2 ---> 18 million reads
T1 ---> 17million reads
T2 --->15million reads
E1 --->16 million reads
E2 --->15 million reads
F1 ---> 12 million reads
F2 --->6 million reads
A1 ---> 17 million reads
A2 ---> 13 million reads

Using the raw counts from these conditions reports only two DEG genes, which I guess because of the huge difference in coverage between replicates. Is there any possible way to use this data to screen significant genes. Any help, suggestions or ideas in this regard would be appreciated.

Cheers

[Ido Tamir]

F2 is the only one that has a low read count compared to the others, but everything else is within a range that the normalisation should handle - and even F2 could be still ok. I don't have any literature reference for this. I guess its rather the high variance between replicates and only 2 replicates per condition. You should look at more diagnostic plots: clustering, PCA, where do the reads align to, ratio of alignments ...

[ATpoint]

Without knowing experimental details such as the differences between the conditions, your experiment is probably underpowered in terms of replicates. n=2 per condition is low. Replication number is more important than sequencing depth.

[vivek37373]

I completely accept your point. However, the sequencing performed was limited to funds.

[Kevin Blighe]

Can you follow up to the points by Ido Tamir, too? - they are quite relevant, in particular, this:

You should look at more diagnostic plots: clustering, PCA, where do the reads align to, ratio of alignments ...