[BioC] Important: Help with eBayes in AgiMicrRna and on how to find differentially expressed genes

[Sean Davis]

On Mon, Feb 6, 2012 at 9:15 AM, Karthik K N <karthikuttan at gmail.com> wrote:
> Dear Sean,
>
> Thank you for your reply. At this time, I am left with no choice but to
> deal with the data as-is. I realize that it is hard to work without
> replicates, but is there any way to make maximum meaningful info from this
> limited data set that I have? Some other tools/algorithms/packages/tests?
>
> Also, what does the contents of ProcessedData.txt file exported from
> AgiMicroRna mean? As far as I know, they are the Total Gene Signals of all
> those miRNAs that pass the filtering step. Isn't it? So, what downstream
> tests can I do using it as an input file? As mentioned in my previous post,
> I am interested in finding out the differentially expressed genes between
> my control and treated samples.

Hi, Karthik.

I'd suggest making a ratio of the treatment versus control.  Those
probes that show a large fold change are the candidates showing
differential expression.  The cutoff for fold-change will be
experiment-specific, so you'll have to look at your data.

Just out of curiosity, what is the reason for the inability to
generate replicates?  If cost is an issue, It is important to realize
that NOT generating replicates can be more costly than actually
generating them, depending on what your followup for candidate
differentially expressed genes will be.

Sean


> On Mon, Feb 6, 2012 at 7:21 PM, Sean Davis <sdavis2 at mail.nih.gov> wrote:
>
>> On Mon, Feb 6, 2012 at 8:29 AM, Karthik K N <karthikuttan at gmail.com>
>> wrote:
>> > Dear Lucas,
>> >
>> > I am using AgiMicroRna package and came across the same error that you
>> also
>> > had i.e., Error in ebayes: No residual degrees of freedom in linear model
>> > fits.
>> >
>> > I read your posts<
>> https://stat.ethz.ch/pipermail/bioconductor/2009-May/027834.html>and
>> > the replies to that.
>> >
>> > My dataset consists of just 2 different arrays, one control and one
>> > treated. I am trying to find out diferentially expressed genes between
>> > them.
>> >
>> > >From the reply to your post I understood the reason why I am seeing this
>> > error with eBayes. I just wanted to ask help from you on how will I be
>> able
>> > to find out differentially expressed genes between these datasets?
>> >
>> > I saw you mentioning BGX, but will that work with Agilent miRNA
>> microarray
>> > platform? My idea is that it will not.
>> >
>> > If there is there anything that you can do to help me solve this, I will
>> be
>> > really thankful to you.
>>
>> Hi, Karthik.
>>
>> The "No residual degrees of freedom" in your case is statistical
>> language meaning that you do not have enough samples to perform the
>> number of tests you want to perform.  In your case, this is a problem
>> because it is not possible to do a statistical hypothesis test with
>> only two samples.  This is not a software problem but a statistical
>> issue.  To "fix" the problem, you will need to perform some biologic
>> replicates.
>>
>> To deal with your data as-is, you can simply compute a ratio between
>> treatment and control.
>>
>> Sean
>>
>
>
>
> --
> Karthik K.N
>
>        [[alternative HTML version deleted]]
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor