Question: No Diffrently Expressed Mirna???
gravatar for Sirus
7.3 years ago by
Sirus790 wrote:

Hi guys, I know this question mqybe look silly but it is taking me a lot of time. I have downoaded some miRNA expression data from the AML samples available on the webstie TCGA (Arround 18 samples). I calsified my samples to groups according to some criteria (aproximatly 4 samples per group), and did a simple diffrent expression analysis sing DESeq Here is the used code:

      library("DESeq") <- data.frame(row.names = colnames( counts ),condition = groups)  
 conds <- factor($condition) 
 cds <- newCountDataSet( counts, conds ) 
 cds <- estimateSizeFactors( cds )
 cds <- estimateDispersions( cds )
 sizeFactors( cds )

 result <- nbinomTest( cds, levels(groups)[1],levels(groups)[2] );
 result = result[order(result$padj), ]

I my data I have 705 miRNA, the problem is after doing the DE analysis I didn't ant DE miRNA (lmfit model didn't converge)

I tried to write my own t-test analysis but I got similar results (when using the Counts, none was DE, but when I used RPKM about 2 or 3 miRNA where DE between the diffrent groups). I used the following code for the t-test and multiple test correction,

get.DiffrentlyExpressedMiRNA<-function(counts, groups){


    Pvals<- rep(0,nrow(counts));
    logFC<- rep(0,nrow(counts));

    grp1<-which(groups == levels(groups)[1]);
    grp2<-which(groups == levels(groups)[2]);

    for(i in 1:nrow(counts)){                
        Pvals[i]<-t.test(counts[i, grp1], counts[i, grp2])$p.value;
        FC[i]<- mean(counts[i,grp1]) / mean(counts[i,grp2]);

    results<-data.frame(miRNA_ID = rownames(counts)[nas], P_val= Pvals[nas], FC = FC[nas], log2FC= logFC[nas], q = qvalue(Pvals[nas])$qvalues);


How do you generally guys, analyse the miRNA data?

NB: I just have access to the level 3 data on TCGA (which means I just have the counts and RPKM for each miRNA)

Any help is apreciated. Thanks,

ADD COMMENTlink modified 7.3 years ago • written 7.3 years ago by Sirus790
gravatar for Jeremy Leipzig
7.3 years ago by
Philadelphia, PA
Jeremy Leipzig19k wrote:

I tend to filter out the miRNAs that don't have enough counts (say 200 total among all samples), since those are only going to mess with correction for multiple testing

ADD COMMENTlink modified 7.3 years ago • written 7.3 years ago by Jeremy Leipzig19k

Hi Jeremy, I removed the weakly expressed miRNA, I got some improvement but still no diffrently expressed miRNA. I think the problem is in the set that I selected. I will check further :)

ADD REPLYlink written 7.3 years ago by Sirus790

another strategy is to normalize the mirna counts against housekeeping genes (are there housekeeping mirnas?) , but i've never tried that.

ADD REPLYlink modified 7.3 years ago • written 7.3 years ago by Jeremy Leipzig19k

Thanks, Jeremy I will take a look at that :)

ADD REPLYlink written 7.3 years ago by Sirus790
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1225 users visited in the last hour