Search
Question: Results with very few genes with qval <0.05
0
gravatar for jdchc
8 months ago by
jdchc0
jdchc0 wrote:

Hello,

I run the new protocol (HISAT, StringTie, Ballgown) with 24 samples (13 unaffected vs 11 affected) to get differential expression. After run R package in Ballgown, I got only 4 genes with a qval under 0.05. My goal is to perform a pathway analysis, but with only 4 genes there is no possible to do that. I was also trying to figure out how ballgown is calculating fold change. Because, I would also like to know which of these are up or down regulated. But I don't know how is the relationship log2(affected/unaffected) or log2(unaffected/affected). When I got the FPKM values and I tried to get the value of log2(a/u), fold change from the results are close but are not the same. My main concern is about qval, I run the same bam file from hisat in cuffdiff and I get more than 3000 significant genes. And even running DESeq from gtf file from stringtie i got around 700 significant genes. For this reason, I don't think quality of the samples could be the problem. Any idea why am I getting only 4 genes?

Thank you 

 

ADD COMMENTlink modified 8 months ago by aiwa0 • written 8 months ago by jdchc0
0
gravatar for kapeelc
8 months ago by
kapeelc0
kapeelc0 wrote:

I have the same issue too. Does it have to do with more stringent value set for FDR? I get comparable results when the FDR is set <0.001 in cudiff. How one sets the FDR in Ballgown?

ADD COMMENTlink written 8 months ago by kapeelc0
This could be for a number of reasons. Could you please make a histogram of the pvalues from each method you are comparing and post them so that I can try to help you diagnose the issue? Best Jeff On Sat, Mar 11, 2017, 7:07 PM kapeelc [bioc] <noreply@bioconductor.org> wrote: > Activity on a post you are following on support.bioconductor.org > > User kapeelc <https: support.bioconductor.org="" u="" 12567=""/> wrote Answer: > Results with very few genes with qval <0.05 > <https: support.bioconductor.org="" p="" 93402="" #93739="">: > > I have the same issue too. Does it have to do with more stringent value > set for FDR? I get comparable results when the FDR is set <0.001 in cudiff. > How one sets the FDR in Ballgown? > ------------------------------ > > Post tags: ballgown > > You may reply via email or visit > A: Results with very few genes with qval <0.05 >
ADD REPLYlink written 8 months ago by Jeff Leek490

Attached are the distribution of pvalues  and qvalues for genes

gene_pvalue_plot

q value plot

ADD REPLYlink written 7 months ago by kapeelc0
0
gravatar for aiwa
8 months ago by
aiwa0
aiwa0 wrote:

Hi, 

I have the same issue. I have RNA seq data for 3 genotypes and 3 cell types. Each cell type has 3 replicates (27 samples in total). When i do pairwise comparisons (one genotype to another for the same cell type), i get very few genes with qvalue less then 0.05. But when i do full model (effect of genotype) then i get many genes with q<0.05. What can be a solution in this case?

ADD COMMENTlink written 8 months ago by aiwa0
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 146 users visited in the last hour