[BioC] siggenes error message about sam.plot2

He, Yiwen (NIH/CIT) [C] heyiwen at mail.nih.gov
Tue Mar 21 18:54:54 CET 2006


Thank you Mr. Schwender,

That really helped!

In case someone is curious how the NAs in d values occur:

Actually there were these warnings when running sam:
1: There are 21084 genes with at least one missing expression value.
The NAs are replaced by the gene-wise mean.
2: 0 of the 21084 genes with at least one NA have no and 28 have one non-missing expression value.
All these 28 genes are removed, and their d-values are set to NA.
3: There are 3 genes with zero variance. These genes are removed,
and their d-values are set to NA.

I looked at my data for those 31 genes:
28 had only 1 value each, 3 had two *identical* values each.

I usually filter out genes with lots of missing values before running sam, but got this dataset at an intermediate state so didn't realize about the missing values.

Thanks again for your help.
Yiwen

-----Original Message-----
From: Holger Schwender [mailto:holger.schw at gmx.de] 
Sent: Tuesday, March 21, 2006 12:23 PM
To: He, Yiwen (NIH/CIT) [C]
Cc: sam-software at yahoogroups.com; bioconductor at stat.math.ethz.ch
Subject: RE: [BioC] siggenes error message about sam.plot2

Hi Yiwen,

> 
> Thank you for the quick response and sorry to make the mistake to email to
> your personal account. I'm wondering, which email list is more appropriate
> for siggenes questions: SAM or BioC?

I would say the BioC mailing list is the more appropriate place for siggenes
questions. 

It is no problem to send emails to my personal accounts but please use this
email address, i.e. holger.schw at gmx.de, since I do not read my emails to
holgerschw at yahoo.de anymore (actually, I thought that I have closed this
account).

To your other questions:

> 
> Yes, my samRestuls at d does have 31 NAs out of 21987 values. 

I have checked it and this is a bug in sam.plot2 which occurs if any of the
d-values is missing.

> Are those NAs
> caused by not having enough replicates in a group? 

Actually, in this case, an error message should occur. I'm not sure why some
of your d-values are missing. Please take a look at the rows of your data
set that correspond to the entries in samResults at d that are NA. Maybe this
gives you an answer to your question.

> Is it true that those
> genes will not be selected for any delta values since the difference
between
> expected and observed will be NA and will not meet the delta threshold?

Genes with a missing d-value are excluded from further analysis. If, e.g.,
31 of your 21,987 genes are NA, only 21,956 (and not 21,987) expected
d-values are computed and so on.

> How
> often does NAs occur in the d values? I have used siggenes to analyze lots
> of datasets and this is the first time I got the sam.plot2 error.
>

This totally depends on your data set. If there are no missing values in
your data set, then actually none of the d-values should become NA. The only
exception I currently can think of is when all the values of a gene are
equal. But in this case you should actually get a warning that there are
genes with zero variance.

Best,
Holger


 
> I appreciate you help!
> Yiwen
> 
> 
> 
> -----Original Message-----
> From: Holger Schwender [mailto:holger.schw at gmx.de] 
> Sent: Tuesday, March 21, 2006 11:33 AM
> To: He, Yiwen (NIH/CIT) [C]
> Cc: sam-software at yahoogroups.com; bioconductor at stat.math.ethz.ch
> Subject: Re: [BioC] siggenes error message about sam.plot2
> 
> Hi Yiwen,
> 
> first of all, sorry that I have not answered your email yet. The reason
> for
> this might be that you have sent your email to my yahoo account which I do
> not use anymore. Please use this email address for further request.
> 
> Now to the second problem: Yes, this bug has been fixed in version 1.3.x
> of
> siggenes.
> 
> To the sam.plot2 problem: I actually do not really know why this happens.
> But it might be a bug in sam.plot2. Please let me know if there is any
> missing value in samResults at d. If so, this might cause the error and I
> will
> fix this bug.
> 
> Best,
> Holger
> 
> 
> 
> 
> > --- Ursprüngliche Nachricht ---
> > Von: "He, Yiwen \(NIH/CIT\) [C]" <heyiwen at mail.nih.gov>
> > An: <sam-software at yahoogroups.com>, <bioconductor at stat.math.ethz.ch>
> > Betreff: [BioC] siggenes error message about sam.plot2
> > Datum: Tue, 21 Mar 2006 10:00:18 -0500
> > 
> > Hi, I tried to reach the maintainer of the siggenes package without much
> > success, so I'm forwarding my question to both lists hoping that some
> > one may have experienced the same problem and know how to fix it. Your
> > help will be highly appreciated!
> > Yiwen
> > 
> > -----Original Message-----
> > From: He, Yiwen (NIH/CIT) [C] 
> > Sent: Thursday, March 16, 2006 4:59 PM
> > To: 'holger schwender'
> > Subject: siggenes error message about sam.plot2
> > 
> > Hello Mr. Holger,
> > 
> > I have been able to use your siggenes package successfully but recently
> > got an error about pos.stats when I tried to make a sam plot for a
> > particular dataset.
> > 
> > dd is a matrix of 21987 x 44, 5 groups:
> > > table(cl)
> > cl
> >  1  2  3  4  5
> >  7 10 10 11  6
> > 
> > > samResults<-sam(dd,cl, B=B, rand=rand.sd)
> > > plot(samResults, 1.6)
> > Error in sam.plot2(x, delta = y, pos.stats = pos.stats, sig.col =
> > sig.col,  :        pos.stats must be either 0 (statistics are not
> > displayed),
> > 1 (stats are displayed in the upper left of the plot), or 2 (lower
> > right).
> > 
> > I got the same message when I call directly:
> > 
> > > sam.plot2(samResults, 1.6)
> > Error in sam.plot2(samResults, 1.6) : pos.stats must be either 0
> > (statistics are not displayed), 1 (stats are displayed in the upper left
> > of the plot), or 2 (lower right).
> > 
> > I was able to make it work by explicitly specify:
> > 
> > > plot(samResults, 1.6, pos.stats=2) OR
> > > sam.plot2(samResults, 1.6, pos.stats=2)
> > 
> > But I don't understand why the internal parameter is not working. 
> > 
> > I also tried to subset the data and interestingly plotting was fine for
> > the first two groups (21987 x 17), but not for the first four groups
> > (21987 x 38). But when trying some other datasets with > 2 groups, it
> > worked fine.
> > 
> > I hope to get some insight from you. I tried both R 2.1/siggenes 1.2.17
> > and R 2.2/siggenes 1.4.
> > 
> > 
> > Also, just to confirm with you, when using R 2.1 and siggenes 1.2.17 for
> > another dataset, I got an error:
> > 
> > > samResults<-sam(dd,cl, B=B, rand=rand.sd)
> > Error in cut.default(s, quan, include.lowest = TRUE, right = FALSE) :
> >         'breaks' are not unique
> > 
> > But when I tried R 2.2 and siggenes 1.4 on the same data, it worked
> > fine. Should I be concerned about my data or was it some bug in siggenes
> > that is fixed in 1.4?
> > 
> > Thank you very much for your help, and I look forward to your answer.
> > 
> > Yiwen He
> > mAdb Team
> > BIMAS/DCB/CIT/NIH
> > 
> > _______________________________________________
> > Bioconductor mailing list
> > Bioconductor at stat.math.ethz.ch
> > https://stat.ethz.ch/mailman/listinfo/bioconductor
> > 
> 
> -- 
> Bis zu 70% Ihrer Onlinekosten sparen: GMX SmartSurfer!
> Kostenlos downloaden: http://www.gmx.net/de/go/smartsurfer
> 

-- 
Echte DSL-Flatrate dauerhaft für 0,- Euro*!
"Feel free" mit GMX DSL! http://www.gmx.net/de/go/dsl



More information about the Bioconductor mailing list