Hey guys,
I just wanted to check to make sure there's nothing else I can do before I give up.
I'm using Tophat to HTSeq to DESeq workflow. I have three conditions, a WT, a knockout, and a control.
After using HTSeq, I copied all my data into one file. I then followed the DESeq vignette. I've attached the plot of my dispersion, my MA plot, and my histogram plot. As you can see, I did not get a single gene that was differentially expressed. In fact, every one of them had a p-adjust value of 1.
I then used the following steps for filtering.
rs<- rowSums (counts(cds))
use <- (rs > quantile(rs, 0.4))
table (use)
FALSE TRUE
11784 17454
cdsFilt <- cds[use, ]
I then re-ran "estimateSizeFactors" and "estimateDispersions". I still did not get a single gene that was differentially expressed. Once again, all of them had a p-adjust value of 1.
I've also attached those plots. Is there anything else I can do? Is there a way to determine if one of my replicates is bad, and that's throwing everything off (I did this in triplicate, with two of my replicates having 50M reads, with the other having 100 M)?
Thanks for any and all advice! I'm pretty close just to giving up, which would be a wasted year, but at least I can just move on or something.
I just wanted to check to make sure there's nothing else I can do before I give up.
I'm using Tophat to HTSeq to DESeq workflow. I have three conditions, a WT, a knockout, and a control.
After using HTSeq, I copied all my data into one file. I then followed the DESeq vignette. I've attached the plot of my dispersion, my MA plot, and my histogram plot. As you can see, I did not get a single gene that was differentially expressed. In fact, every one of them had a p-adjust value of 1.
I then used the following steps for filtering.
rs<- rowSums (counts(cds))
use <- (rs > quantile(rs, 0.4))
table (use)
FALSE TRUE
11784 17454
cdsFilt <- cds[use, ]
I then re-ran "estimateSizeFactors" and "estimateDispersions". I still did not get a single gene that was differentially expressed. Once again, all of them had a p-adjust value of 1.
I've also attached those plots. Is there anything else I can do? Is there a way to determine if one of my replicates is bad, and that's throwing everything off (I did this in triplicate, with two of my replicates having 50M reads, with the other having 100 M)?
Thanks for any and all advice! I'm pretty close just to giving up, which would be a wasted year, but at least I can just move on or something.
Comment