[BioC] [Fwd: problem : Detection-p.value column in Illumina HumanHT12 data]

Md.Mamunur Rashid mamunur.rashid at kcl.ac.uk
Wed Sep 9 15:39:20 CEST 2009


-------- Original Message --------
Subject: 	Re: Creating LumiBatch object from ExpressionSet
Date: 	Wed, 9 Sep 2009 13:53:42 +0100
From: 	Md.Mamunur Rashid <mamunur.rashid at kcl.ac.uk>
To: 	Pan Du <dupan at northwestern.edu>, "bioconductor at stat.math.ethz.ch" 
<bioconductor at stat.math.ethz.ch>



Dear Dr. Pan Du,

Thanks for your reply.I am looking at the gene filter package. Hope I can find out a way around.
The data I have, does contain detection column (column names : "Detection - 4765967022_A"......).
I am using lumiR("file-name.txt",detectionTh = 0.01)function to read the file. But the

>  pData(featureData(norm_object))$presentCount always return null

Is there anything special needed to be done to bring the detection information to the lumi Object??

I have also tried :
data_raw<- lumiR("Aneurist_Sample_Probe_Profile_Modified.txt",convertNuID=FALSE,inputAnnotation=FALSE, columnNameGrepPattern = list(exprs='AVG_SIGNAL', se.exprs='BEAD_STD', detection='Detection', beadNum='Avg_NBEADS'))


I have a attached a snapshot of the data with the email ("showing the detection column"). please have a look.

*** This is a Illumina HumanHT12 chip data (produced by bead-studio) with 96 samples.
While pre-processing the raw data I only have the following warning:

" There is no control probe information in the LumiBatch object
   No background adjustment will be performed.
   done.
"
Do you think this might have any significance in the above context.???


Thanks in advance for your valuable help.

regards,
Md.Mamunur Rashid



On 09/08/2009 04:21 PM, Pan Du wrote:
> Hi Mamun
>
> For Illumina microarray, the detection p-value basically represents 
> whether the probe intensity is significant higher than the background 
> level (estimated based on the distribution of the negative control 
> probes.). When the p-value is lower than some threshold, for example, 
> 0.05, we can claim this probe is “Present”. If no measurement is 
> “Present” for the probe, we usually consider it is not interest in our 
> analysis and can be prefiltered. Since your data has no detection 
> p-value information, you have to use other prefiltering method, such 
> as those defined in genefilter package.
>
> For your second question, I am not sure what exactly your question is. 
> I read your attached document. It looks fine for me.
>
>
> Pan
>
>
> On 9/7/09 9:34 AM, "Md.Mamunur Rashid" <mamunur.rashid at kcl.ac.uk> wrote:
>
>     Dear Dr. Pan Du,
>     Thanks for your response. Please accept my apology for replying so
>     late. The object AP seems  a numeric list of all the genes
>     detection values
>     (mean from all samples).Can you please give a bit further
>     elaboration about
>     how "AP" can be used to extract the genes expressed under p.value
>     threshold ???
>
>     More over a have few more questions. I would be really helpful if
>     you can give
>     me some suggestions regarding the problems.
>
>
>     I am working with a set of illumina microarray data (96 samples)
>     from three groups
>     (i.e. group-1(X),group-2(Y),group-3(Z)). 32 samples from each group.
>
>     I have read the data using lumiR method and processed the data
>     using lumi Methods.(lumiE)
>     Now I need to identify the differentially expressed genes by
>     comparing each of these groups
>     with each other. I am using linear model in limma package and
>     topTable method to identify
>     top N differentially expressed genes.(please have a look at the
>     attachments).
>
>     ***I have attached a file containing (1. The code, 2. result of
>     the top-table, 3. Array weights
>
>     Thanks in advance.
>
>     regards,
>     mamun
>
>
>
>
>
>     Sincerely,
>     Md.Mamunur Rashid
>
>
>
>     On 07/31/2009 05:16 PM, Pan Du wrote:
>
>
>         Hi Md.Mamunur,
>
>         Functions lumiQ and detectionCall does not support
>         ExpressionSet objects
>         because they need some information not included in
>         ExpressionSet class by
>         default. For example, both detectionCall and lumiQ function
>         need to use the
>         data matrix included in the "detection".
>
>         If you ExpressionSet object do include "detection", here is
>         the alternative
>         way to estimate number of detection calls of each probe:
>
>         detect <- get('detection', assayData(lumi.T))
>         ## suppose you claim p-value less than 0.05 as a detection call
>         AP <- rowSums(detect <= 0.05)
>
>         Hope this is helpful to you,
>
>
>         Pan
>
>



More information about the Bioconductor mailing list