SEQanswers

Go Back   SEQanswers > Applications Forums > RNA Sequencing



Similar Threads
Thread Thread Starter Forum Replies Last Post
DESeq: defining conditions mouchkam Bioinformatics 2 04-10-2012 05:45 PM
Is more than two conditions possible in DESEQ? greener RNA Sequencing 5 05-09-2011 04:10 PM
RNASeq experiments with 2 conditions and more than 3 replicates Jouneau Luc RNA Sequencing 1 04-05-2011 04:00 AM
PCR conditions for PE adapters Luke Illumina/Solexa 3 03-02-2011 09:48 AM

Reply
 
Thread Tools
Old 05-21-2012, 12:54 PM   #1
yylilly
Junior Member
 
Location: US

Join Date: May 2012
Posts: 2
Smile DEXSeq - more than two different conditions

I am still working on preparing the data for running DEXSeq. I have a question. Will DEXSeq work if there are more than two conditions, e.g. cond1,cond2,cond3.. Thanks!
yylilly is offline   Reply With Quote
Old 05-22-2012, 01:07 AM   #2
areyes
Senior Member
 
Location: Heidelberg

Join Date: Aug 2010
Posts: 165
Default

Yes it will. If you run an analysis with all your conditions, it will show those exons in which at least in one condition is different from the other conditions.
areyes is offline   Reply With Quote
Old 06-25-2013, 01:25 PM   #3
sdm
Junior Member
 
Location: Cambridge, UK

Join Date: Oct 2009
Posts: 9
Default DEXSeq with single-end reads

Is it possible to do a high-quality differential splicing analysis based on single-end reads. Or does it have to be paired-end data. Any views on this are highly appreciated
sdm is offline   Reply With Quote
Old 06-26-2013, 12:02 AM   #4
areyes
Senior Member
 
Location: Heidelberg

Join Date: Aug 2010
Posts: 165
Default

Sure, at least with DEXSeq s it is possible.
areyes is offline   Reply With Quote
Old 06-26-2013, 12:04 AM   #5
sdm
Junior Member
 
Location: Cambridge, UK

Join Date: Oct 2009
Posts: 9
Default

Quote:
Originally Posted by areyes View Post
Sure, at least with DEXSeq s it is possible.
Thanks for your reply. Are you actually one of the developers of DEXSeq?
sdm is offline   Reply With Quote
Old 06-26-2013, 12:06 AM   #6
sdm
Junior Member
 
Location: Cambridge, UK

Join Date: Oct 2009
Posts: 9
Default

Quote:
Originally Posted by sdm View Post
Thanks for your reply. Are you actually one of the developers of DEXSeq?
Sorry another question, maybe you can point me to a study/paper, where it has been done by single-end sequencing?
sdm is offline   Reply With Quote
Old 06-26-2013, 12:11 AM   #7
areyes
Senior Member
 
Location: Heidelberg

Join Date: Aug 2010
Posts: 165
Default

The original paper has both single-end and paired-end data:

http://www.ncbi.nlm.nih.gov/pubmed/22722343
areyes is offline   Reply With Quote
Old 07-01-2013, 01:18 PM   #8
h_manoj
Junior Member
 
Location: San Diego

Join Date: Oct 2012
Posts: 6
Default DEXSeq multiple tissues

Quote:
Originally Posted by areyes View Post
Yes it will. If you run an analysis with all your conditions, it will show those exons in which at least in one condition is different from the other conditions.

Hello,
I have 37 RNA-seq datasets for 18 human tissues - some have 3 replicates, some 2 and some do not have any (sorry to say that).

I'm thinking of a way to do differential exon usage analysis - using DEXseq.

I could not see any example in the documentation where multiple conditions are compared (eg., tissues) other than pairwise comparisons. In edgeR, I was able to use the ANOVA-like test for getting differentially expressed genes (across an average of all 18 tissues). I was wondering if there is such a test I could do with DESeq/DEXseq.

Thanks,
Manoj.
h_manoj is offline   Reply With Quote
Old 02-13-2014, 01:58 AM   #9
Jane M
Senior Member
 
Location: Paris

Join Date: Aug 2011
Posts: 239
Default

Quote:
Originally Posted by areyes View Post
Yes it will. If you run an analysis with all your conditions, it will show those exons in which at least in one condition is different from the other conditions.
Hello everybody,

I have RNA-Seq paired-ends data for 3 conditions: Control (4 biological replicates), Leukemic cells with mutation in gene X (6 biological replicates) and Leukemic cells without mutation in gene X (4 biological replicates).

One of the questions I try to answer is the differential exon usage induced by this mutation. I am using DEXSeq_1.8.0.

Basically, I need to compare exon level in leukemic cells with and without the mutation.
Since I have additional information with the control, I would like to output the genes that show a differential expression in at least one exon between both conditions leukemic cells with and without the mutation, with in addition the level for the controls.
To be clear, I don't want to see the genes that show a differential expression in at least one exon between conditions leukemic cells with or without the mutation and the control.

Is it possible with DEXSeq?
If not, I have 2 choices: either I "loose" the information from the control condition or I see all the changes including those that do not come from the mutation (that's what I did for now).

Thank you in advance,
Jane
Jane M is offline   Reply With Quote
Old 02-13-2014, 04:24 AM   #10
areyes
Senior Member
 
Location: Heidelberg

Join Date: Aug 2010
Posts: 165
Default

Hi @Jane M,

You could try to change the GLM formulas to include all the variables of your analysis. First, I would recommend updating your R and DEXSeq version. if your design data frame looks like this:

Code:
phenotype   mutatedGene
control   no
control   no
control   no
control   no
leukemia   yes
leukemia   yes
leukemia   yes
leukemia   yes
leukemia   yes
leukemia   yes
leukemia   no
leukemia   no
leukemia   no
leukemia   no
you could try to include the formulas:

Code:
formulaFullModel <- ~ sample + exon + phenotype:exon + mutatedGene:exon
formulaReducedModel <- ~ sample + exon + phenotype:exon
And proceed as in the section "Additional technical or experimental variables" of the vignette
areyes is offline   Reply With Quote
Old 02-13-2014, 06:10 AM   #11
Jane M
Senior Member
 
Location: Paris

Join Date: Aug 2011
Posts: 239
Default

Quote:
Originally Posted by areyes View Post
you could try to include the formulas:

Code:
formulaFullModel <- ~ sample + exon + phenotype:exon + mutatedGene:exon
formulaReducedModel <- ~ sample + exon + phenotype:exon
Yes, good idea! I will try within next days. I did not think about my problem this way, thank you!
Jane M is offline   Reply With Quote
Old 02-14-2014, 04:23 AM   #12
areyes
Senior Member
 
Location: Heidelberg

Join Date: Aug 2010
Posts: 165
Default

Thinking about it again, I am not sure what will be the output of the testing. To add the "phenotype" as a blocking factor we would also need samples with phenotype "control" and mutatedGene "yes". Of course this is not possible in this experiment! How do the "plotDEXSeq" plots look like?

Maybe would be good to try to use the "formulaFullModel" to estimate dispersions, and the testing doing ignoring the "phenotype" variable...

Let me know what comes out!
areyes is offline   Reply With Quote
Old 03-14-2014, 04:42 AM   #13
Jose Garcia
Junior Member
 
Location: Milan

Join Date: Jul 2013
Posts: 4
Default full formula

Hi Alejandro,
I have one condition with :
CTRL
CTRL
CTRL
CTRL
H24
H24
H48
H48

Can I use the defalut full formula?
Where can I found exons for H24-CTRL and H48-CTRL as result?
Thanks
Jose
Jose Garcia is offline   Reply With Quote
Old 03-14-2014, 05:25 AM   #14
areyes
Senior Member
 
Location: Heidelberg

Join Date: Aug 2010
Posts: 165
Default

Hi Jose,

The default formula would test for differences in exon usage in any of the conditions.

If you are interested in the specific changes between H24-CTRL and between H48-CTRL it is better to subset the data accordingly and run the analysis separately.

Alejandro
areyes is offline   Reply With Quote
Old 03-14-2014, 05:34 AM   #15
Jose Garcia
Junior Member
 
Location: Milan

Join Date: Jul 2013
Posts: 4
Default

Thanks Alejandro,
I thought something similar to to what Mike uses in DESeq2 could be used in DEXSeq but with exons behind is harder.
I'll subset the two and do it separately.

Jose

Quote:
Originally Posted by areyes View Post
Hi Jose,

The default formula would test for differences in exon usage in any of the conditions.

If you are interested in the specific changes between H24-CTRL and between H48-CTRL it is better to subset the data accordingly and run the analysis separately.

Alejandro
Jose Garcia is offline   Reply With Quote
Old 03-28-2014, 03:22 PM   #16
xrao
Junior Member
 
Location: Houston USA

Join Date: Mar 2014
Posts: 9
Default

Quote:
Originally Posted by areyes View Post
Hi Jose,

The default formula would test for differences in exon usage in any of the conditions.

If you are interested in the specific changes between H24-CTRL and between H48-CTRL it is better to subset the data accordingly and run the analysis separately.

Alejandro

Hi, Alejandro

I have a similar situation. One thing that I am a little worried is, when seperating the data into 2 (part 1 with H24 and control, and part 2 with H48 and control) and do whole DEXSeq analyses seperately, the adjusted p-values may not have been adjusted enough. The BH correction may need to be applied to all the p-values of the tests from the 2 analyses. I am not sure if I am right or not on this?

Thanks,
Xiayu
xrao is offline   Reply With Quote
Old 05-14-2014, 01:52 PM   #17
GreboGuru
Junior Member
 
Location: Maine, USA

Join Date: Jul 2013
Posts: 7
Default

Hello all,
I am having a similar problem to the posts above. I have been using edgeR for gene level expression differences and need help translating the contrasts I have been using in edgeR over to a factorial model in DEXseq.

In edgeR I have combined all of my experimental factors into one, for example:
Sample Group
Sample1 Drug.2h
Sample2 Drug.0h
Sample3 Placebo.2h
Sample4 Placebo.0h
ect for replicates...

and contrast setup is nearly identical to one described in the edgeR vignette:
(Drug.2h-Drug.0h)-(Placebo.2h-Placebo.0h)

However I am at a loss at how do to this comparison using a factorial model in DEXseq.
Should I break up my factors like so:

Sample Treat Time
1 Sample1 Placebo 0h
2 Sample2 Placebo 0h
5 Sample5 Placebo 2h
6 Sample6 Placebo 2h
7 Sample1 Drug 0h
8 Sample2 Drug 0h
11 Sample5 Drug 2h
12 Sample6 Drug 2h


and then use the following factorial models?
formulaFullModel = sample + exon + treatment:exon + time:exon + treatment:time:exon
formulaReducedModel = sample + exon + treatment:exon + time:exon

Thanks ahead of time for your consideration.
GreboGuru is offline   Reply With Quote
Old 05-16-2014, 03:05 AM   #18
areyes
Senior Member
 
Location: Heidelberg

Join Date: Aug 2010
Posts: 165
Default

Hi @GreboGuru,

I think that looks reasonable, you would be testing for those cases where the treatment has a specific effect in exon usage in a specific time point.

Alejandro
areyes is offline   Reply With Quote
Old 05-16-2014, 07:31 AM   #19
GreboGuru
Junior Member
 
Location: Maine, USA

Join Date: Jul 2013
Posts: 7
Default

Alejandro,
Hello back and thank you for your feedback, it is invaluable. Running the analysis now.
GreboGuru is offline   Reply With Quote
Old 05-17-2014, 01:52 PM   #20
josephyan
Junior Member
 
Location: NYC, U.S.

Join Date: May 2011
Posts: 2
Default

See its paper: http://genome.cshlp.org/content/22/10/2008.full
Quote:
Note that we test against the null hypothesis that none of the conditions influences exon usage, and hence, if there are more than two different conditions ρ, we aim to reject the null hypothesis already if any one of the conditions causes differential exon usage.
So I think the answer is Yes
josephyan is offline   Reply With Quote
Reply

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 07:31 PM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2019, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO