[BioC] Different gcrma output values after upgrade BioC 2.0 -> 2.4

[Rullmann, J.A.C. (Ton)]

Hi Jean,
>From your answer I get the idea that only probes with (very) low
intensity should be affected. However, I find differences for probes at
all levels of intensity (2log intensity after gcrma between 2 and 15).

In the HGU133PLUS2 example given in my first email the difference
between the BioC 2.0 and 2.4 results from gcrma are not that large: over
48 arrays the ratio new/old of the log-intensity values is 1.001 +/-
0.020, and the ratio of the intensity values is 1.007 +/- 0.090.
 
In another set of arrays, of the RGU34A type, the difference between the
BioC 2.0 and 2.4 results from gcrma is much larger: over 8 arrays the
ratio new/old of the log-intensity values is 1.05 +/- 0.16, and the
ratio of the intensity values is 1.4 +/- 1.3!
For ~5 % of the cases the new intensity value is more than four-fold
different than the old value. The effects are partially conserved
between arrays, but identification of regulated genes still gives a list
that is 10% different than the previously obtained set, using the same
selection criteria.
Differences already occur after bg.adjust.gcrma. I verified by dumping
the data from the R sessions that the probe files that were used
(rgu34aprobe_2.4.0 and rgu34aprobe_2.2.0) are identical.

Does anyone else have similar experiences? What is the explanation?

Thanks for your help. 

Ton Rullmann
T: +31 412 661423
ton.rullmann at spcorp.com


-----Original Message-----
From: Zhijin Wu [mailto:zwu at stat.brown.edu] 
Sent: Thursday, 27 August, 2009 15:21
To: Rullmann, J.A.C. (Ton)
Cc: bioconductor at stat.math.ethz.ch
Subject: Re: [BioC] Different gcrma output values after upgrade BioC 2.0
-> 2.4

The most recent change is on probes with intensity at lower bounds. 
These use to go through the same adjustment for specific binding 
(GSB.adjust), but are no longer adjusted b/c they are believed to bear 
no signal.

-- thanks to Wei Keat Lim and colleagues who brought it up in their 
paper "Comparative analysis of microarray normalization procedures:
effects on reverse engineering gene networks"
Lim et al, Bioinformatics 2007 Vol23(13):i282

Rullmann, J.A.C. (Ton) wrote:
> Hi,
> We upgraded from BioC 2.0 to BioC 2.4. Now I note that log-intensities
> computed by gcrma differ from the values I got previously, for the
same
> input.
> For example, for one array the mean ratio of old vs new
log-intensities
> is 0.9996 with an sd of 0.0187.
> For a second array, analyzed in the same set, the mean and sd values
are
> 0.9991 and 0.0197.
> This implies there are dozens of probe sets for which the computed
> log-intensity changes by ~ 10%.
> 
> Is this "expected behaviour", that is: has the gcrma algorithm
changed?
> 
> Thanks,
> 
> Ton Rullmann
> Senior Research Scientist
> Project Leader Systems Biology
> Dept. Molecular Design and Informatics
> Schering-Plough
> RK 2309
> PO Box 20
> 5340 BH Oss, The Netherlands
> www.schering-plough.com
> 
> 
> 
> This message and any attachments are solely for the intended
recipient. If you are not the intended recipient, disclosure, copying,
use or distribution of the information included in this message is
prohibited --- Please immediately and permanently delete.
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives:
http://news.gmane.org/gmane.science.biology.informatics.conductor


-- 
------------------------------------
Zhijin (Jean) Wu
Assistant Professor of Biostatistics
Brown University, Box G-S121
Providence, RI  02912

Tel: 401 863 1230
Fax: 401 863 9182
http://www.stat.brown.edu/zwu


This message and any attachments are solely for the intended recipient. If you are not the intended recipient, disclosure, copying, use or distribution of the information included in this message is prohibited --- Please immediately and permanently delete.