Is there a way to calculate p-values, FDR or t test starting from logFC data?

Question

[Spacebio]

I am new in DEA and I have RNA-Seq data already normalised to TPM, I calculated the log2FC values but I would like to calculate the p-values, FDR or t-values. Is there a way to do this kind of analysis in R? Here people discuss about it but the data is normalised in a different way and they didn't really answer the question. Any help is very appreciated.

[EagleEye]

Hi Spacebio,

Do not make redundant posts with similar question. I found this thread started by you TPM to logFC and pvalues

[Spacebio]

Hi EagleEye, I'm sorry! I am learning to use this platform to ask questions or debate about some topics. I couldn't find my old question, for that reason I started a new one. Please be a bit patient, slowly but surely I am learning to handle the info here.

Answer 1

[Spacebio]

I found this question so I guess that is a helpful way to do so.

d <- DGEList(counts = myDF[,2:7], group = c(rep("S1", 3), rep("S2", 3)), genes = myDF[,1])
d <- calcNormFactors(d)
d <- estimateCommonDisp(d)
d <- estimateTagwiseDisp(d)
d <- estimateTrendedDisp(d)
de <- exactTest(d, pair=c("S1", "S2"), dispersion = "tagwise")
deDF <- as.data.frame(de)

Any other ideas? I heard there is a way to convert TPM to raw counts, is that true? Any help is very appreciated!

Answer 2

[TriS]

it's a (partial) duplicate question from here: select DE genes here How To Calculate Differential Gene Expression In Rnaseq Experiments?

see answers there.

also from Google:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4728800/

https://www.bioconductor.org/help/workflows/rnaseqGene/

http://biocluster.ucr.edu/~rkaundal/workshops/R_mar2016/RNAseq.html

https://web.stanford.edu/class/bios221/labs/rnaseq/lab_4_rnaseq.html

there is a minimum knowledge or R required.

otherwise even Galaxy seems to have a workflow

[Spacebio]

Hello, thank you for your help and suggestions! The problem here is that I have a .csv file containing RNA-Seq values normalised already to TPM, I have no access to the raw data. That is the reason why I am performing the analysis by hand. I calculated the log2FC values but now I would like to calculate the p-values, FDR or t-test in order to make sure that my results make sense. I am using R, but most of the packages ask for the raw data. Any other suggestion?

[TriS]

you don't need raw count values, you can use TPM and limma to get DEGs

Answer 3

[theobroma22]

You can use different R packages like QValue available in Bioconductor, or R functions like the p-value and t-test function. Their usages are pretty straightforward, but if you have issues just start a new post so we can try to help you.

[Spacebio]

Hello, thanks for your suggestion. The problem here is that R packages ask for the raw data and I have no access to it. My data is a .csv file containing RNA-SEQ data normalised already to TPM.

[theobroma22]

Typically after normalization comes differential expression analysis. This is done by developing a model, linear or other. So, since you have normalized values now you can start to compare your null and full models. From this comparison you can get your p-values, q-values and FDR. In summary, you don't have to start with the raw data to finish / complete your analysis.

[Spacebio]

I didn't know that! Thank you very much for the info, I am new in this kind of analysis and there is a lot of info I am missing.