Entering edit mode
3.1 years ago
sabaghianamir70 ▴ 60
I having a problem with analyzing my data.. I have 5 time points of 5 patients, So the problem is, when i take all timepoints 0 from all patients and compare them with timepoints 1 from all patients, my p value adj failed. P.s i cant compare each person separately, because i just have one sample per each timepoint, So i at least need two replicate of each sample to do DEG.
Any suggestions are appropriated. Thanks.
May I ask why the experiment was designed in this way? Which differential expression analysis program are you using?
Deseq and edgeR. i tried both. As i said : I thought i can compare all time 0 from 5 patients to all time 1.
Thanks, but you are contradicting yourself with these statements:
Can you please clarify all samples that you have, and what is the specific issue? Is the issue just that no gene passes FDR correction? In that case, perhaps there genuinely is no statistically significant difference. You could try to reduce the 'FDR burden' by eliminating low count genes prior to normalisation, if not already done.
Explaining your experimental design and conditions may also further help us to understand any other potential issues. For example, you may need to adjust for confounding factors or other sources of bias.
I have 5 patients which all have 5 timepoints.(0 1 3 5 7) So i just need to take all samples which belongs to time 0, and compare them to other time points For example : 0-1 0-3 0-5 0-7 The main issue is about my FDR and p value adj, because in when i run it with edgeR, i have no significant up or down. Also i have MDS plot like that which showing some thing is wrong. My datas are Mir-seq and after cleaning i used bowtie and featureCount for further steps. I hope these information are what you want to know. Thanks a lot for helping :)
Please show full code and how you set up the design matrix and contrasts.
because I should pay for using cloud services, i just trying to test the whole thing in galaxy. If it helping, i can just send all workflow here.
Thanks, but just the design matrix, please.
this is how i set up my analysis..each Count file includes 5 time points.
Ok, we are at a dead-end here. The problem that me and others were asking about is how the tool knows which samples belong to one group. This seems to be a blackbox since you are using Galaxy. Therefore, you have two options. First, see if you can get the raw count matrix and then do analysis locally, 100% following edgeR manual. Second, set
TRUEand then upload it somewhere, then please share the link. Then we can inspect what Galaxy is doing.
Yeah, well this does not help at all. If you cannot show code it is difficult/impossible to debug. Things to check are if experimental groups are properly set, if design matrix is correct and if the contrasts are properly set up, and how multiple testing correction was done. Get raw data and follow the edgeR manual, it covers all you need for standard analysis.
Could you be more specific about why p-value adjustment "failed"? Is it just that you didn't observe any significant changes.
As you can see in this picture, All p value adj are above the 0.05
Ah. So they haven't failed, they're just not significant. Did you control for the different baseline expression level in your 5 patients?
How can i check that ? I almost tried every option in edgR or Deseq2. can you explain how can i control the different baseline expression level ?