problem with no replicate data analysis

Question

[Bioinformatics1]

Help!

I have no replicate data. I spent a lot of time doing downregulated/upregulated and then do Pathway Analysis.

Method 1: Rsubread and the edgeR quasi-likelihood pipeline: it does not give me a Dispersion estimation. As a result, I can not continue to calculate res <- glmQLFTest(fit, contrast=B.LvsP), neither pathway analysis

Method 2: DESEQ2 R does not calculate dds. Link: https://bioconductor.org/packages/release/workflows/vignettes/RnaSeqGeneEdgeRQL/inst/doc/edgeRQL.html

Method 3: Cannot draw voom plot and gives error fit.cont <- eBayes(fit.cont). " no residual". Link: https://combine-australia.github.io/RNAseq-R/06-rnaseq-day1.html#Testing_relative_to_a_threshold_(TREAT)

Method 4: Page 24 gives a solution for no replication data. I do not know what is housekeeping index is and what I should add to my R pipeline

 y0 <- estimateDisp(y1[housekeeping,], trend="none", tagwise=FALSE)
Error in h(simpleError(msg, call)) : 
  error in evaluating the argument 'i' in selecting a method for function '[': object 'housekeeping' not found

Link: https://www.bioconductor.org/packages/devel/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf

Method 5: it is not easy to follow. Link:https://www.rstudio.com/wp-content/uploads/2015/02/data-wrangling-cheatsheet.pdf

I appreciate it if you help me by example or by way to follow comments

Thanks in advance

Answer 1

[Gordon Smyth]

The edgeR User's Guide (to which you link) has an entire section called "What to do if you have no replicates". That section enumerates all the feasible possibilities.

[Bioinformatics1]

Gordon, What do I add instead of the housekeeping index? I only have data, I don't know how to add the housekeeping gene. Please read page 24

[Gordon Smyth]

Actually I wrote page 24 of the edgeR User's Guide, but I can't give you customized advice specific to your dataset. If you know nothing at all about your own data, for example which genes should not be DE, then you need to use a different method.

[Bioinformatics1]

it is my pleasure to communicate with you. Do you have an example of no replicate data for RNA seq analysis to share with me? RNA seq lab sent me the information of samples in fastq format

[ATpoint]

Just download any published dataset that has two conditions and then use a single replicate per condition if you need practice data. Or go for the airway dataset at Bioconductor and then extract a single sample per condition. But you have data it seems, why do you need data "shared" with you? Do you know how to process fastq files? If not check Salmon for RNA-seq preprocessing.

[Bioinformatics1]

Hi ATpoint, I uploaded and did alignment in base space Illumina. How I can find housekeeping index?

I have data but when I go through the pipeline, I get errors for dispersion v <- voom(y,design,plot = TRUE)

for below equation gives me an error : "no residual" fit.cont <- eBayes(fit.cont) and I can not run this command because of previous eBayes error summa.fit <- decideTests(fit.cont)

And I can not continue to find Upregulated and downregulated.

For Pathway analysis I need to use "fit.cont " too but eBayes gave me the error and it does not let me continue according to pipeline. I wanted to use Gordon's instruction ( no replicate). I do not understand what information or command I replace it with instead of "housekeeping" index in the equation. So I asked him to provide an example if it is possible and helped me to move forward and complete my project.

[ATpoint]

Why don't you simply use approach 2 in this "No replicate" section of the manual so "making up" a dispersion estimate? It does not really matter because it is equally unreliable and arbitrary as all other suggested steps but at least the most simple to follow and you get some preliminary results you can look at. Keep in mind that such as approach is neither reliable nor publishable, and only useful for careful results exploration.

[Bioinformatics1]

Does Approach 2 mean" DESEQ2 R"? it gives an error.

in Method 4 (page 24) I need DGEList for continuing the pipeline on my data. It starts from dispersion calculation.

bcv <- 0.2 counts <- matrix( rnbinom(40,size=1/bcv^2,mu=10), 20,2) y <- DGEList(counts=counts, group=1:2) et <- exactTest(y, dispersion=bcv^2)

[Carlo Yague]

housekeeping index is the row number of the genes in the dataset that are supposed to not be differentialy expressed in your condition. You have to make an educated guess in choosing these, which require knowledge of your organism and experimental conditions.

[Bioinformatics1]

Hi Carlo, I used metformin on human cells. Is it possible to guide me more?

Thanks a lot