[BioC] DexSeq extracols option

[Ioannis Vlachos]

Hello,

One small tip that I found out the hard way.

In DEXSeqHTML the results table [  results <- DEUresultTable(ecs)] is
rounded to the third digit prior to padj filtering.

Therefore, this might create discrepancies in nrows between the extraCols
data.frame and the result  table.

So the recipe for a successful extraCols data.frame has been changed to:

Get the results table-> round to the third digit-> filter with desired
p-value (or q value)-> get the required info from biomaRt or wherever-> keep
unique lines only and sort accordingly -> use as ecols dataframe

Cheers.

IV

-----Original Message-----
From: Alejandro Reyes [mailto:alejandro.reyes at embl.de] 
Sent: Monday, July 29, 2013 10:42 AM
To: Ioannis Vlachos
Cc: bioconductor at r-project.org
Subject: Re: [BioC] DexSeq extracols option

Dear Ioannis,

Thanks for your message, it is usually nice and easier to track the problem
if you add the output of your sessionInfo and reproducible code, but I am
guessing that you are using a old version of DEXSeq.

The table that the HTML reporter produces a table with information for each
gene, not for each exon.  I would be almost sure that the 396 are the number
of genes that have significant exons.  As the documentation indicates, the
parameter extraCols accepts a data frame in which the rownames are genes of
your ExonCountSet object.

Let me know if this is the problem,

Best regards,
Alejandro


> Hello,
>
> A (hopefully) simple question.
>
> As I saw in ??DEXSeqHTML, there is an option to add attributes in the 
> HTML results (extracols).
>
> I downloaded from biomaRt a set aof attributes to test this (gene 
> name, biotype and description), and created a data frame having as a 
> first column the exons of my ecs  (retaining exon order), geneID and the
new attributes.
>
> The first two columns were taken by accessing fData from the ecs.
>
>   The resulting data frame naturally had an equal number of rows with 
> the exons present in my ecs object (355320).
>
> I tried to set the extracols option in a different number of ways 
> (putting as row.names the geneID or exonID, having only one column 
> etc), but I always got this error:
>
> Error in data.frame(..., check.names = FALSE) :
>    arguments imply differing number of rows: 355320, 396
>
> Where 355320 is the correct number of rows of both ecs exons and of my 
> extra attributes data.frame. (I haven't figured out what has 396 rows
yet).
>
> I checked the online reference manual and there is no mention of the 
> extracols option. I also checked the code of  the relevant function 
> and I don't see it get used somewhere.
>
> Any ideas?
>
> Is there another way to add gene or exon attributes in the HTML?
>
> Thank you all,
>
> Best regards,
>
> Ioannis
>