[BioC] sva/ComBat with missing values?

[Johnson, William Evan]

Hollis, 

What are you using? ComBat or SVA? 

ComBat does work with some missingness but its it depends on what is missing. For example, if you are missing all values for one batch for one gene, you will get a singularity error.

Evan



On Sep 24, 2013, at 6:00 AM, <bioconductor-request at r-project.org>
 wrote:

> Date: Mon, 23 Sep 2013 20:32:22 -0400
> From: Jeff Leek <jtleek at gmail.com>
> To: Hollis Wright <wrighth at ohsu.edu>
> Cc: "bioconductor at r-project.org" <bioconductor at r-project.org>
> Subject: Re: [BioC] sva/ComBat with missing values?
> Message-ID:
> 	<CAGWgrqMC9gyU4JuiVjtgXT01_6-rLAVfO1ix=mmq0AKPhJxYzg at mail.gmail.com>
> Content-Type: text/plain
> 
> Hi Hollis
> 
> At the moment the only way you can perform these analyses is on a complete
> matrix of genomic measurements. Depending on the structure of the missing
> data, one approach is to use the impute package:
> 
> http://www.bioconductor.org/packages/release/bioc/html/impute.html
> 
> to impute the missing values before performing Combat/SVA.
> 
> Hope this helps!
> 
> Jeff
> 
> 
> 
> 
> On Mon, Sep 23, 2013 at 4:08 PM, Hollis Wright <wrighth at ohsu.edu> wrote:
> 
>> Hi, all; I have a 224-gene OpenArray RT-PCR data set that I'm analyzing,
>> and I strongly suspect one of the runs is having a batch effect on the
>> entire experiment from EDA and analysis of the quality values.
>> Unfortunately, this means that there are a number of missing values in the
>> matrix (though not all from the suspect run). If I pull the genes
>> containing these values, SVA/ComBat works nicely to clean up the data that
>> remains; if I don't, it fails with a "system is exactly singular" Lapack
>> routine error. Is there any way to force these functions to use something
>> like the "pairwise.complete.obs" from cor to work around the missing
>> values, or is my only option to drop genes?
>> 
>> Hollis Wright, PhD
>> Ojeda Lab, Division of Neuroscience
>> Oregon Health and Science University
>> _______________________________________________
>> 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
>> 
>