[BioC] 回复： TMM factors

[Mark Robinson]

Dear Sooby,

Please keep posts on the list and see:
http://bioconductor.org/help/mailing-list/posting-guide/

On 21.04.2012, at 18:17, Sooby wrote:

> I'm not sure about the library size you mentioned.
> Is it the total reads of a sample or the sum of the reads mapped to each gene.
> I mean, if the total reads of a sample is 1,000,100,
> but when I mapped the reads to the gene, I got a list of gene with reads mapped to it, and the sum of this is 900,000.
> Do I use 1,000,000 or 900,000??

I feel like I answered this already.  My previous response was "Usually, total number of mapped reads.", which would suggest 900,000 in your example.

> The normalization factor I calculate is like 0.5,1.5,1.3 ,but in you manual they are all near to 1.  
> Is it right??

It is more extreme than I normally see, but it could be ok.  Have you looked at your data?  Figure 1 of the edgeR user's guide gives an example "smear" plot.  Perhaps you should look at one of those for your data and assess whether the normalization factors are appropriate.

Best,
Mark


> 发件人： Mark Robinson <mark.robinson at imls.uzh.ch>
> 收件人： Sooby <sooby24 at yahoo.cn> 
> 抄送： "bioconductor at r-project.org" <bioconductor at r-project.org> 
> 发送日期： 2012年4月21日, 星期六, 上午 5:20
> 主题: Re: [BioC] TMM factors
> 
> Hi Sooby,
> 
> Some comments below.
> 
> On 20.04.2012, at 04:32, Sooby wrote:
> 
> > Hi all,
> > I met a question when I analysed mRNA data. I use TMM provided in edgeR to normalizition my data. After I got the factors, How Do I treat my raw counts?? In my oppion, (raw.counts)*10^7/(librarysize*factors),Is It right?
> 
> So, this is correct, for a reads per 10M conversion.  I'm not sure what you mean by "treat my raw counts".  If you are doing differential expression with edgeR (or DESeq, etc.), we strongly suggest to use the raw counts, without "treating" them.
> 
> 
> > And another question is which librarysize do I need? The unique total reads or the reads mapped to each gene???
> 
> Usually, total number of mapped reads.
> 
> Best,
> Mark
> 
> > I need your help~~~
> > Best 
> > 
> >     [[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
> 
> 
>