[BioC] edgeR : How to associated trait with DE?

Gordon K Smyth smyth at wehi.EDU.AU
Thu Sep 22 01:01:17 CEST 2011


Dear Sermsawat,

Technical replicates should be aggregated by summing, not by taking the 
mean.

Best wishes
Gordon

---------------------------------------------
Professor Gordon K Smyth,
Bioinformatics Division,
Walter and Eliza Hall Institute of Medical Research,
1G Royal Parade, Parkville, Vic 3052, Australia.
Tel: (03) 9345 2326, Fax (03) 9347 0852,
smyth at wehi.edu.au
http://www.wehi.edu.au
http://www.statsci.org/smyth

On Wed, 21 Sep 2011, Sermsawat Tunlaya-anukit wrote:

> Thank you very much for code example. some traits i measure and has 
> technical replicate can i assign those data into design metrics or i 
> just use mean.
>
> Best regards,
> Sermsawat T.

> On Sep 15, 2011, at 1:22 AM, Gordon K Smyth wrote:
>
>> Dear Sermsawat,
>>
>> NAs are not permitted in design matrices in edgeR.  If you have NAs, 
>> you have to remove these libraries explicitly:
>>
>>  i <- !is.na(Growth64)
>>  disph64 <- estimateGLMCommonDisp(d[,i], design[i,])
>>
>> Best wishes
>> Gordon
>>
>>
>> On Wed, 14 Sep 2011, Sermsawat Tunlaya-anukit wrote:
>>
>>> Thank you for your recommend. I also have some trait which has missing 
>>> data (NA). Do edgeR estimateGLMcommonDisp has any option that skip 
>>> this missing data? I try na.rm=TRUE but it isn't help.
>>>
>>>
>>>> design <- model.matrix(~group,data=d$samples)
>>>> designh64 <- cbind(design,Growth64)
>>>> designh64
>>>       (Intercept) group2 group3 group4 Growth64
>>> Rwt_2             1      0      0      0       48
>>> Rwt_3             1      0      0      0       NA
>>> Rwt_4             1      0      0      0       NA
>>> RL4_3_1           1      1      0      0       32
>>> RL4_4_2           1      1      0      0       15
>>> RL4_5_1           1      1      0      0       22
>>> RL4_6_2           1      1      0      0       25
>>> RL7_1_2           1      0      1      0       15
>>> RL7_2_3           1      0      1      0       10
>>> RL7_3_1           1      0      1      0       12
>>> RL8_3_1           1      0      0      1       12
>>> RL8_3_2           1      0      0      1       12
>>>> disph64<-estimateGLMCommonDisp(d,designh64)
>>> Error in qr.default(X) : NA/NaN/Inf in foreign function call (arg 1)
>>>>
>>>> disph64<-estimateGLMCommonDisp(d,designh64, na.rm=TRUE)
>>> Error in dispCoxReid(y, design, offset = offset, ...) :
>>> unused argument(s) (na.rm = TRUE)
>>>
>>> Sermsawat Tun.
>>>
>>>
>>> On Sep 14, 2011, at 7:01 PM, Gordon K Smyth wrote:
>>>
>>>> Dear Sermsawat,
>>>>
>>>> In edgeR, you can simply add a quantitative column to the design 
>>>> matrix, then use glmFit() and glmLRT() to test for the significance 
>>>> of the coefficient for that column.
>>>>
>>>> I do not recommend that you do the calculations using the pseudo 
>>>> counts.
>>>>
>>>> Best wishes
>>>> Gordon
>>>>
>>>>> Date: Tue, 13 Sep 2011 23:01:29 -0400
>>>>> From: Sermsawat Tunlaya-anukit <stunlay at ncsu.edu>
>>>>> To: bioconductor at r-project.org
>>>>> Subject: [BioC] edgeR : How to associated trait with DE?
>>>>>
>>>>> Can i associated trait into edgeR result? I have quantitative traits 
>>>>> such as height in each library that would like to find gene that 
>>>>> co-regulate or co-expression with that trait. How can i apply this 
>>>>> into edgeR? Can i use $psedo.alt to calculate correlation with my 
>>>>> trait? Or i should GLM to calculate DE that associated with trait? I 
>>>>> am not sure how to apply my trait into design or another vector and 
>>>>> calculate estimateGLMCommonDisp.
>>>>>
>>>>> Thank in advance,
>>>>> Sermsawat Tunlaya-anukit

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}



More information about the Bioconductor mailing list