DESeq question

[chrisbala]

Hi Simon (and others),

Quick question about DESeq. So far it looks like a nice approach! But it was giving me some somewhat strange (although possible reasonable) numbers...so I wanted to check....

Sometimes the values estimated for baseMeanA and baseMeanB differ in the direction and the extent of change from the original raw counts. Is this expected? These baseMeans are meant to be 'corrected' estimates of read counts for the conditions in question? If I am getting such differences between the raw read counts and the baseMeans, is this simply a reflection of the strength of the dataset (I am running without biological reps). If so, thats fine (I think). But just making sure all is well..

For example...

Counts_Novel	Counts_Familiar	DESeq-id	DESeq-baseMean	DESeq-baseMeanA	DESeq-baseMeanB	DESeq-foldChange
141	139	28	140.176915	133.0658319	147.2879981	1.106880677
143	136	30	139.5312055	134.9532905	144.1091204	1.067844436

Is this normal?

[Simon Anders]

Hi Chris

If, by "strength", you mean sequencing depth, yes.

In order to make the samples comparable, differences in sequencing depth (a.k.a. library size) have to be addressed. If sample A is sequenced 20% deeper than sample B, then 120 counts in sample A should be considered as equal in expression strength to 100 counts in B.

The function 'estimateSizeFactors' estimates the sequencing depth, and with 'sizeFactors', you can print out the estimates. In the example just given, you might get size factors 1.2 and 1.0 for A and B (or another pair of numbers with the ratio 1.2:1).

To get the 'baseMean' value, all count values are transformed to the "base scale" (or "common scale" in the paper) by dividing them with the appropriate size factor.

I've explained our scheme to estimate the size factors a while ago in this post.

Simon

[chrisbala]

counts deseq

hi simon,

nope, thats not what i meant by 'strength', but it makes much much more sense...

thanks!

[m!x]

Hi Simon,

I am trying to install DESeq package on Mac OS X 10.4 and Bioconductor 2.10.
But I keep getting this error message:

$ R CMD install /Users/lilyanamargaretha/Desktop/DESeq
* installing to library ‘/Library/Frameworks/R.framework/Resources/library’
* installing *source* package ‘DESeq’ ...
** libs
** arch - i386
gcc -arch i386 -std=gnu99 -I/Library/Frameworks/R.framework/Resources/include -I/Library/Frameworks/R.framework/Resources/include/i386 -I/usr/local/include -fPIC -g -O2 -c pval.c -o pval.o
cc1: error: unrecognized command line option "-arch"
make: *** [pval.o] Error 1
ERROR: compilation failed for package ‘DESeq’
* removing ‘/Library/Frameworks/R.framework/Resources/library/DESeq’
* restoring previous ‘/Library/Frameworks/R.framework/Resources/library/DESeq’

Any idea how to fix it?

Thanks!

[Simon Anders]

Hi M!x

Hi Simon,
I am trying to install DESeq package on Mac OS X 10.4 and Bioconductor 2.10.
But I keep getting this error message:

[...]
cc1: error: unrecognized command line option "-arch"

This is a most peculiar error. Your C compiler does not accept the '-arch' option. However, this is a standard option of GCC (and, if I am not completely confused, has been so since long before thre release of Tiger). Hence, something in your Xcode installation seems to be utterly broken.

Could you please type in a terminal the commands "gcc -v" and "which gcc" and send me the output (by private mail to anders(at)embl(dot)de, in order to not clog the forum with details)?

Cheers
Simon

[m!x]

Hi Simon,

I finally managed to install DESeq package and ran it pretty well.
Thanks!

[ikumar2]

Hi Simon,

I am trying to use DESeq. After step-

cds <- newCountDataSet( countsTable, conds)

I am getting this error-

Error in round(countData) : Non-numeric argument to mathematical function

Any suggestions? Thanks in advance.

[Simon Anders]

cds <- newCountDataSet( countsTable, conds)

I am getting this error-

Error in round(countData) : Non-numeric argument to mathematical function

It probably means that your countsTable contains non-numeric data, e.g. string with gene names. It should contain only numbers. Make sure to cut away any columns with annotation data and/or move them to the row / column names.

Simon

[dgu]

DESeq installation doubts...

Hi M!x



This is a most peculiar error. Your C compiler does not accept the '-arch' option. However, this is a standard option of GCC (and, if I am not completely confused, has been so since long before thre release of Tiger). Hence, something in your Xcode installation seems to be utterly broken.

Could you please type in a terminal the commands "gcc -v" and "which gcc" and send me the output (by private mail to anders(at)embl(dot)de, in order to not clog the forum with details)?

Cheers
Simon

Hello Simon!

Hope you can give me a hand about your program.

I started a session in R 2.10.1 and typed the following to load DESeq:

> source("http://www.bioconductor.org/biocLite.R")
> biocLite("DESeq")
Using R version 2.10.1, biocinstall version 2.5.10.
Installing Bioconductor version 2.5 packages:
[1] "DESeq"
Please wait...

Warning message:
In getDependencies(pkgs, dependencies, available, lib) :
package ‘DESeq’ is not available

Do you recognize the warning message? I think it may have not installed DESeq after all, 'cause R doesn't seem recognize the function "newCountDataSet".

Sorry if this is too basic but just started with R and DESeq. Any pointers will be appreciated.

Daniel

[Simon Anders]

I started a session in R 2.10.1 and typed the following to load DESeq:

> source("http://www.bioconductor.org/biocLite.R")
> biocLite("DESeq")
Using R version 2.10.1, biocinstall version 2.5.10.
Installing Bioconductor version 2.5 packages:
[1] "DESeq"
Please wait...

Warning message:
In getDependencies(pkgs, dependencies, available, lib) :
package ‘DESeq’ is not available

Do you recognize the warning message? I think it may have not installed DESeq after all, 'cause R doesn't seem recognize the function "newCountDataSet".

Sorry if this is too basic but just started with R and DESeq. Any pointers will be appreciated

It says: "package ‘DESeq’ is not available". So, of course, you haven't installed it. The Bioconductor web server had occasional hick-ups recently, so maybe it really was not available for a while. Please try again, and, if it still does not work, follow the manual installation instructions here.

Simon

[Wolfgang Huber]

Bioconductor versions for DESeq

Hi dgu and Simon,

DESeq was introduced into Bioconductor at release 2.6. You are trying with release 2.5, that's why you get "package ‘DESeq’ is not available". Bioconductor releases are synchronised with R releases, so Bioconductor 2.6 goes with R 2.11. The recommended approach is to install R >= 2.11 and try with "biocLite" again, as you described. See also http://www.bioconductor.org -> Getting Started.

Nothing to do with server availability, which as far as I can tell has been fine.

Wolfgang Huber

[dgu]

Hi dgu and Simon,

DESeq was introduced into Bioconductor at release 2.6. You are trying with release 2.5, that's why you get "package ‘DESeq’ is not available". Bioconductor releases are synchronised with R releases, so Bioconductor 2.6 goes with R 2.11. The recommended approach is to install R >= 2.11 and try with "biocLite" again, as you described. See also http://www.bioconductor.org -> Getting Started.

Nothing to do with server availability, which as far as I can tell has been fine.

Wolfgang Huber

Thanks, Wolfgang!

Indeed, I downloaded R 2.11.1 and the problem was taking care of. Thanks!

[dgu]

It says: "package ‘DESeq’ is not available". So, of course, you haven't installed it. The Bioconductor web server had occasional hick-ups recently, so maybe it really was not available for a while. Please try again, and, if it still does not work, follow the manual installation instructions here.

Simon

Hi Simon!

Thanks for your reply. I sorted the problem out using the newest version of R.

I am going through each step of differential gene expression analysis from the DESeq manual. It is great and very well explained.

I had one question about Fig. 2 (pg 7). I managed to get the "smoothScatter" function to work, but the regression fit (in red) doesn't want to come out. Below is the script:

## Fit from the local regression
lines(log10(fittedBaseVar)~log10(baseMean),diagForYC[order(diagForYC$baseMean),],col="red")

The plotting windows kinda freezes indefinitely and I get the following message:

Error in plot.xy(xy.coords(x, y), type = type, ...) :
plot.new has not been called yet

I've rewritten the script many times with no success. Do you see anything wrong?

Thanks in advance for any pointers!

Daniel

[Simon Anders]

Hi Daniel,

maybe you have closed the plotting windows after calling 'smoothScatter'? The 'lines' function expects that a plot is already displayed so that the axes ares set.

Simon

[dgu]

Thanks for the prompt answer, Simon!

Indeed... that was the problem. I was closing the window.

R is great. It is just not THAT straightforward, if you know what I mean.

Thanks again for your help. Do you have any updates on your DESeq paper?

Cheers,
Daniel