Hi everyone,
I'm running into a problem where the second biological replicates have significantly less IP efficiency and reads than the first replicate, so when I tried running the DESeq2 analysis to find differential enrichment analysis, there were no significant differences between the two dataset even though there are known differences in the literatures.
An example is as such: where the read counts for a certain peak for the first replicate (A_1, B_1) are both higher than the read counts for the second replicate (A_2, B_2) but the over all trend if we are only comparing the replicate itself is that A > B. However because A_2 is lower than B_1 this cause the DESeq2 to not see this as different.
A_1 1200 A_2. 669 B_1. 756 B_2. 240
I also tried normalizing the read counts by library sizes, read in peaks but none of them seem to fix the problem as when I did a heatmap or PCA, A_1 will always correspond better to B_1 and A_2 to B_2 because A_1 and B_1 have more reads than A_2 and B_2.
Is there any other way I can approach this problem in terms of normalization? or doing DE on the dataset? Thank you.
DESeq2's built-in normalization method accounts for differences in overall sequencing depth among samples. If you try using the
plotCountsfunction on your dds object for a few of your genes of interest, you can see what the normalized read counts look like for different samples and experimental groups. That might be useful.okay thanks for the tip! I think after plotting the counts and seeing how it compares to the normalized count, I'm more confused on why my DESeq2 showed 0 differential binding sites. Because it seems like before normalizing, the counts are very close to each other between the two groups but when the counts are normalized, you can see the differences between the group but its not reported in DESeq2.