Differentially expressed genes

Parharn · 02-24-2014, 08:44 AM

Hi,
I am doing RNA-seq analysis using Tuxedo pipeline. I wonder when do you name it DIFFERENTIALY expressed genes? Does it have to be significant genes out of cuffdiff analysis or any sort of fold changes between two conditions.
I am asking it because my two conditions show very few significant changes in gene expression. And that is between G2 and S phase cells I just get 16 significant differentialy expressed genes.
Let me know if you need further information.

thanks.

Bukowski · 02-24-2014, 02:02 PM

Significance in cuffdiff is when the q-value (FDR adjusted p-value) is less than the p-value itself and it is not dependent on the magnitude of the fold change.

(Source, the cuffdiff manual : http://cufflinks.cbcb.umd.edu/manual.html#gene_exp_diff )

It's impossible to tell why you have few significant genes without more information about your experimental set up. Too few reads per sample, or too few replicates would be possible explanations.

Parharn · 02-24-2014, 02:16 PM

Yes, I understood how Cuffdiff defines significant, but my question is if differentially expressed genes are specifically those that are expressed significantly?

Second, since we were establishing RNA-seq for the first time in our lab, we did only one replicate of each condition. We just wanted to go all the way to the end before doing actual experiment. I recently read that number of replicates have great impact on the accuracy of data analysis. My data has between 10 to 20 million reads per sample. I guess having one replicate is making the problem. What do you think?

How many replicates do you suggest with how many reads in each?
Please let me know what other information will make it more clear.

Thanks.

TiborNagy · 02-26-2014, 12:27 AM

Great article about the replicates/read numbers
http://www.ncbi.nlm.nih.gov/pubmed/24319002

Jeremy · 02-26-2014, 09:20 PM

Most qPCR papers use a minimum fold change of 2 and statistical significance before calling anything differentially expressed. Since RNA-seq is basically high throughput qPCR it would make sense to use the same thresholds.

reema · 02-27-2014, 03:03 AM

How about using logFold change?

Parharn · 02-27-2014, 08:04 AM

Thanks people.

Krish_143 · 02-27-2014, 01:07 PM

When we have no replicates.. then i suggest u to try Gfold.
Gfold > 0 || gfold < 0 means it Upregulated or down regulated.

For more information check.
http://compbio.tongji.edu.cn/~fengjx/GFOLD/gfold.html

Parharn · 02-27-2014, 03:45 PM

Thanks a lot Krish. Looks very interesting, how can I get the software? Apt-get I guess? I am new to Linux environment.

Krish_143 · 02-27-2014, 11:30 PM

You can download at
https://bitbucket.org/feeldead/gfold/downloads

For downlading to linux :

wget https://bitbucket.org/feeldead/gfold....V1.1.1.tar.gz

Check Readme and installation files in gfold.

Parharn · 02-27-2014, 11:56 PM

Thanks a million Krish. I will give it a try.

Parharn · 03-04-2014, 03:16 AM

This Gfold was a game changing program. Thanks for suggesting it Krish.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Cufflinks, differentially expressed genes	statsteam	Bioinformatics	5	11-15-2013 12:28 PM
still too many differentially expressed genes.	thejustpark	Bioinformatics	2	11-03-2013 02:54 PM
What to do after finding differentially expressed genes?	sazz	Bioinformatics	1	07-15-2013 11:39 AM
very different numbers of differentially expressed genes by DESeq	bliu1	RNA Sequencing	2	08-22-2012 09:41 AM
DESeq and EdgeR: too many differentially expressed genes!?!?	cutcopy11	Bioinformatics	5	12-08-2011 01:14 AM

02-24-2014, 08:44 AM	#1
Parharn Member Location: Sweden Join Date: Jul 2013 Posts: 84	Differentially expressed genes Hi, I am doing RNA-seq analysis using Tuxedo pipeline. I wonder when do you name it DIFFERENTIALY expressed genes? Does it have to be significant genes out of cuffdiff analysis or any sort of fold changes between two conditions. I am asking it because my two conditions show very few significant changes in gene expression. And that is between G2 and S phase cells I just get 16 significant differentialy expressed genes. Let me know if you need further information. thanks.

02-24-2014, 02:02 PM	#2
Bukowski Senior Member Location: UK Join Date: Jan 2010 Posts: 390	Significance in cuffdiff is when the q-value (FDR adjusted p-value) is less than the p-value itself and it is not dependent on the magnitude of the fold change. (Source, the cuffdiff manual : http://cufflinks.cbcb.umd.edu/manual.html#gene_exp_diff ) It's impossible to tell why you have few significant genes without more information about your experimental set up. Too few reads per sample, or too few replicates would be possible explanations. Last edited by Bukowski; 02-24-2014 at 02:04 PM. Reason: added link

02-27-2014, 01:07 PM	#8
Krish_143 Member Location: Sweden Join Date: Jan 2012 Posts: 45	When we have no replicates.. then i suggest u to try Gfold. Gfold > 0 \|\| gfold < 0 means it Upregulated or down regulated. For more information check. http://compbio.tongji.edu.cn/~fengjx/GFOLD/gfold.html __________________ Krishna

02-27-2014, 11:30 PM	#10
Krish_143 Member Location: Sweden Join Date: Jan 2012 Posts: 45	You can download at https://bitbucket.org/feeldead/gfold/downloads For downlading to linux : wget https://bitbucket.org/feeldead/gfold....V1.1.1.tar.gz Check Readme and installation files in gfold. __________________ Krishna Last edited by Krish_143; 02-27-2014 at 11:35 PM.

02-24-2014, 02:16 PM	#3
Parharn Member Location: Sweden Join Date: Jul 2013 Posts: 84	Yes, I understood how Cuffdiff defines significant, but my question is if differentially expressed genes are specifically those that are expressed significantly? Second, since we were establishing RNA-seq for the first time in our lab, we did only one replicate of each condition. We just wanted to go all the way to the end before doing actual experiment. I recently read that number of replicates have great impact on the accuracy of data analysis. My data has between 10 to 20 million reads per sample. I guess having one replicate is making the problem. What do you think? How many replicates do you suggest with how many reads in each? Please let me know what other information will make it more clear. Thanks.

02-26-2014, 12:27 AM	#4
TiborNagy Senior Member Location: Budapest Join Date: Mar 2010 Posts: 329	Great article about the replicates/read numbers http://www.ncbi.nlm.nih.gov/pubmed/24319002

02-26-2014, 09:20 PM	#5
Jeremy Senior Member Location: Pathum Thani, Thailand Join Date: Nov 2009 Posts: 190	Most qPCR papers use a minimum fold change of 2 and statistical significance before calling anything differentially expressed. Since RNA-seq is basically high throughput qPCR it would make sense to use the same thresholds.

02-27-2014, 03:03 AM	#6
reema Member Location: Scotland Join Date: Feb 2014 Posts: 27	How about using logFold change?

02-27-2014, 08:04 AM	#7
Parharn Member Location: Sweden Join Date: Jul 2013 Posts: 84	Thanks people.

02-27-2014, 03:45 PM	#9
Parharn Member Location: Sweden Join Date: Jul 2013 Posts: 84	Thanks a lot Krish. Looks very interesting, how can I get the software? Apt-get I guess? I am new to Linux environment.

02-27-2014, 11:56 PM	#11
Parharn Member Location: Sweden Join Date: Jul 2013 Posts: 84	Thanks a million Krish. I will give it a try.

03-04-2014, 03:16 AM	#12
Parharn Member Location: Sweden Join Date: Jul 2013 Posts: 84	This Gfold was a game changing program. Thanks for suggesting it Krish.