[BioC] How to save result from limma

[Jenny Drnevich]

Hi Gordon,

I guess a sort="none" would be better - I was just thinking about 
what I have to do, and not what would make the most sense in the 
function! One advantage of the output from topTable over the fitted 
model object is that it contains adjusted p-values while the other 
only contains the raw p-values. I do different manipulations with the 
topTable results depending on what my clients want, but almost 
invariably I back calculate to regular fold-changes and remove the t 
and B columns at the very least. Some want separate "up" and "down" 
lists, some want me to put "up" and "down" in the output, some want 
expression levels added, etc.

Cheers,
Jenny

At 06:07 PM 8/12/2008, Gordon K Smyth wrote:
>Hi Jenny,
>
>It would seem a waste to resort toptable by names ... that would 
>essentially amount to sorting the table, and then unsorting it 
>again. Something like sort="none" would be better.  Also given that 
>limma handles data from all sorts of microarray platforms, it always 
>can't be relied on that the rownames are gene names.
>
>My comment about toptable not being the right function, was on the 
>assumption that anyone creating a toptable of all probes was likely 
>to write it out to a file.
>
>If you find topTable a useful way to get summary statistics for the 
>whole arrays, that's fine.  What sort fo manipulations do you do 
>with the toptable in R?  I guess my thought was that most 
>calculations in R would be done with the fitted model object directly.
>
>Best wishes
>Gordon
>
>On Tue, 12 Aug 2008, Jenny Drnevich wrote:
>
>>Yeah, now I remember that being in the help page for topTable(). 
>>However, I'm not sure why it's not the "right" function. Assuming 
>>all the arguments are equivalent, the values are the same as those 
>>output from write.fit(). I like topTable() better because it gives 
>>you both the raw p-values and the adjusted P-values. Maybe it's 
>>wrong function because it will give you the genes in different 
>>orders for each coefficient? I prefer it to write.fit so I can work 
>>with the values in R without having to read the output file back 
>>in! However, I do have to re-sort it if I want to use with other 
>>objects I've created. I've always wanted a resort.by option of 
>>"gene" that would do my resorting automatically:
>>
>>>allgenes <- topTable(fit,number=nrow(fit))
>>
>>>allgenes <- allgenes[order(as.numeric(rownames(allgenes))),]
>>
>>Any chance of you adding that in the future, Gordon? If not, maybe 
>>you can expand upon why topTable is not the "right" function.
>>
>>Thanks!
>>Jenny
>>
>>At 02:47 PM 8/12/2008, James W. MacDonald wrote:
>>>Actually, write.fit() is what you are looking for. From ?topTable:
>>>Note:
>>>
>>>      This is not the right function to use to create summary statistics
>>>      for all the probes on an array. Please consider using 'write.fit'
>>>      or 'write' for this purpose, rather than using 'topTable' with
>>>      'number=nrow(fit)'.
>>>;-P
>>>Jenny Drnevich wrote:
>>>>Hi Jixin,
>>>>You're mixing up the functions topTable() and write.table() and 
>>>>their arguments. Try this:
>>>>  > allgenes <- topTable(fit, number= nrow(fit), adjust="BH")
>>>>  > write.table( allgenes, file= "la.txt")
>>>>HTH,
>>>>Jenny
>>>
>>>Jenny Drnevich, Ph.D.
>>>
>>>Functional Genomics Bioinformatics Specialist
>>>W.M. Keck Center for Comparative and Functional Genomics
>>>Roy J. Carver Biotechnology Center
>>>University of Illinois, Urbana-Champaign
>>>
>>>330 ERML
>>>1201 W. Gregory Dr.
>>>Urbana, IL 61801
>>>USA
>>>
>>>ph: 217-244-7355
>>>fax: 217-265-5066
>>>e-mail: drnevich at illinois.edu