[BioC] Re: Bioconductor Digest, Vol 11, Issue 12

[Stephen P. Baker]

Bioconductor Digest, Vol 11, Issue 12I would think that ANY significant differential expression by ANOVA is evidence of presence.  Also if using MAS5 or LiWong PM-MM, any significant effect including intercept would be evidence of presence but I don't think the same can be said of RMA or PM only estimates.

-.- -.. .---- .--. ..-.
Stephen P. Baker, MScPH , PhD(ABD)                      (508) 856-2625
Senior Biostatistician                                                     (775) 254-4885 fax
Information Services Bioinformatics Unit
Lecturer in Biostatistics , Graduate School of Biomedical Sciences
University of Massachusetts Medical School
55 Lake Avenue North                          stephen.baker at umassmed.edu
Worcester, MA 01655  USA




Message: 2 
Date: Fri, 09 Jan 2004 17:00:07 -0500 
From: Naomi Altman <naomi at stat.psu.edu> 
Subject: Re: [BioC] P calls (VSN and RMA) 
To: w.huber at dkfz-heidelberg.de, Isaac Neuhaus <isaac.neuhaus at bms.com> 
Cc: rafa at jhu.edu, Petra B Ross-MacDonald 
        <Petra.RossMacDonald at bms.com>,  robert.nadon at mcgill.ca, 
        bioconductor at stat.math.ethz.ch 
Message-ID: <6.0.0.22.2.20040109165122.01dbfcc8 at stat.psu.edu> 
Content-Type: text/plain; charset="us-ascii"; format=flowed 

I have also been working on this problem. 

I compared the Affy "present" calls, and calls based on various levels of 
normalized expression.  Needless to say, these do not match well. 

In our study, it is known that some genes do express at very low levels in 
one of our conditions, and do not express under the other 
conditions.  These genes were declared "not present" in all conditions both 
by Affy and by our (admittedly arbitrary) cut point  (which was 50). 

I did a gene-by-gene ANOVA (which included all genes, even if 
"absent").  Interestingly enough, a few of these genes had a statistically 
significant ANOVA F-test and a look at the expression values confirmed that 
this was due to much higher expression  values (2-fold or more) in the 
known condition.  This seemed to me to indicate that perhaps we ought to 
consider lowering the cut point.  However, if we do this, we also include a 
lot more genes that appear (by RT-PCR) to really be absent. 

So, now I wonder if I can use the ANOVA to provide information about when a 
gene is present.  I appreciate this discussion, because it is an important 
issue for the group of biologists I work with. 

--Naomi 

	[[alternative HTML version deleted]]