To get p-values for the TPM

Question

[VITALA]

I generated TPM for the RNAseq data that am dealing with ,as the samples that i have pooled for my project are from different experiments and my requirement is to get genes exclusive to one condition irrespective of high/low expresion, i generated TPM from RSEM tool, the RSEM output didn't give any p-value as oppposed to what was mentioned in their manual. I need p-values . I understood that feeding RSEM output /raw counts to the Deseq2 is recommended through tximport , but not the normalised counts. Now can anyone suggest how to get p-values for this data? what are the statistical tests that i should perform? I know this must have been asked multiple times on this platform.Still I had to ask this , as i don't have anyother way but to generate the TPM .Kindly help me

Answer 1

[dsull]

There are plenty of RSEM -> tximport -> DESeq2 tutorials online. Just follow one of those.

DESeq2 will give you "p values" based on a Wald statistical test performed on the parameters estimated from the DESeq2 model. If this is confusing, first, understand what p-values are. P-values are used in hypothesis testing. If you have treatment samples vs. control samples, you can test a hypothesis of whether there are differences between the two groups. That hypothesis testing is what gives you a p-value metric.

Saying you want p-values from RSEM output is meaningless. Why do you want p-values? What are you testing for? Obtaining quantifications (like p-values) and slapping them onto a paper without understanding what they are or what you're testing for is a terrible idea.

[VITALA]

sir, deseq2 takes counts as inputs rightfor the hypothesis testing /to see the variance between the two groups, but in my case what i want is to proceed further with TPM, due to my specific requirement, so, i am asking this question.limma trend helps in differential FPKM analysis , by log2 transformation, but again it is not recommended, so i wanted to ask if i can use the same for my TPM analysis also. RSEM manual mentions that 'sample_name.stat/sample_name.pval_LL should have the p value within the sample_name.stat folder, but i didn't get any p value stats in my output folder .Thus, the problem Thankyou

[dsull]

Your post should be a "com, not an "answer". Use "Add Comment" to request clarifications on an answer someone has provided.

tximport can take in the RSEM output files -- I really don't understand why you don't just use that... RSEM output gives you everything you need for DESeq2 -- estimated counts, TPMs, gene lengths, etc. and can be imported directly into DESeq2.

Do you even know what the RSEM p-value does? If so, then why do you even want to use it? What are you trying to do with it?

[ATpoint]

Seconding that. Just use established pipelines, tximport => DESeq2, get your stats and call it a day. Custom pipelines paired with little experience in such analysis is almost always prone to produce wonky results. You can still calculate TPMs from the DESeq2 counts and use that for whatever downstream visualization you want, but be sure that the stats are produced correctly starting from the raw gene level counts as returned by tximport starting from rsem.

[swbarnes2]

RSEM returns a kind of p-value if you run it with the "prior-enhanced" option. Did you do that? Are you completely sure that's the value you want, and that you don't want what DESeq does?

[VITALA]

yes i want what rsem does and i did try with prior-enhanced option/pRSEM, but i think pRSEM is for chipseq data

[dsull]

OK, not sure why your intention is to use RSEM p-values for, but at least tell us the commands you ran then otherwise how are we supposed to help you diagnose the issue?