Go Back   SEQanswers > Applications Forums > RNA Sequencing

Similar Threads
Thread Thread Starter Forum Replies Last Post
statisctics for common genes form DESeq2-edgeR-Cuffdiff Marys RNA Sequencing 0 05-19-2015 07:03 AM
glm edgeR issue giuliano stirparo Bioinformatics 1 01-01-2014 04:48 PM
DESeq2 glm caveats lukas1848 Bioinformatics 2 12-06-2013 08:41 AM
The road after DESeq2 vs edgeR vs Cuffdiff results sindrle Bioinformatics 0 10-26-2013 02:34 AM
edgeR - Run GLM on only one gene nnn RNA Sequencing 1 05-07-2012 01:42 AM

Thread Tools
Old 10-12-2017, 05:02 PM   #1
Location: Montana

Join Date: Nov 2011
Posts: 23
Default Inconsistent gene lists: DESeq2, EdgeR, CuffDiff, and homemade glm

In an effort to get a reliable list of DEGs, I ran CuffDiff, EdgeR, and DESeq2. I though I would grab the overlapping genes and be good to go. However, the DEG lists outputted by each were quite different. Because of this, we decided to also write a GLMM that would be more transparent and easy to understand. This GLMM also gave a fairly different list (with some overlap of course).

The group I'm working with is inclined to go with the GLMM since we (the statistician in the group) knows what it is doing. My main worry is the CuffLinks, EdgeR, and DESeq2 are making some corrections to account for the biology of RNAseq data that we may not understand or incorporate into our GLM.

What important ways do CuffLinks, EdgeR, and DESeq2 diverge from a GLM? We're using glmer.nb with offset=log(librarysize).

Thank you,

We have 14 samples done in triplicates.
Read coverage is ranges from 16-55 million 100bp PE reads per sample.
Aligned to mm10 with Tophat in cufflinks. 83-91% reads mapped.
Used tophat alignments with HTSeq-count to get tables for DESeq2, EdgeR, and GLMM.
Did 7 pairwise contrasts between treatments.

I attached a couple of images.

The venn diagrams show the overlap in DEG lists outputted by the 4 methods for the 7 different contrasts.

Second image shows MC plots for each contrast. X-axis is log2 of the mean htseq counts for all the genes. Y-axis is log2 fold change of treatment vs. mock.
Attached Images
File Type: jpeg venn_all_log0.5_cpm5_small.jpeg (83.8 KB, 8 views)
File Type: jpg Rplot10_mcplots_all_small.jpg (70.8 KB, 9 views)
jvanleuven is offline   Reply With Quote
Old 10-12-2017, 05:29 PM   #2
Senior Member
Location: East Coast USA

Join Date: Feb 2008
Posts: 6,800

This is not an unheard of problem. There have been papers comparing different methods for DE and the results they produce. Here is one of them. There is at least one more that I remember but can't find at the moment.

Last edited by GenoMax; 10-12-2017 at 05:31 PM.
GenoMax is offline   Reply With Quote
Old 10-12-2017, 11:40 PM   #3
Location: Montana

Join Date: Nov 2011
Posts: 23

Thank you for the paper. Below are several more that address the issue.
I'm still not sure if we should pick a method or use some agglomeration of them.

Seyednasrollah F, Laiho A, Elo LL. 2015. Comparison of software packages for detecting differential expression in RNA-seq studies. Brief Bioinform. 16:5970. doi: 10.1093/bib/bbt086.

Zhang ZH et al. 2014. A Comparative Study of Techniques for Differential Expression Analysis on RNA-Seq Data. PLOS ONE. 9:e103207. doi: 10.1371/journal.pone.0103207.

Rapaport F et al. 2013. Comprehensive evaluation of differential gene expression analysis methods for RNA-seq data. Genome Biology. 14:R95. doi: 10.1186/gb-2013-14-9-r95.
jvanleuven is offline   Reply With Quote

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

All times are GMT -8. The time now is 12:04 AM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2018, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO