[BioC] E. coli problems: bimodality and gcrma

[Jenny Drnevich]

Hello everyone,

I am starting to analyze a dataset from Affymetrix's E.coli genome 2.0 
arrays, and I'm running into several weird things. This is the first time 
I've handled prokaryotic data, so I don't know if that's the reason it 
looks so different from most eukaryotic data I've seen. I would really 
appreciate it if anyone with E. coli experience would be willing to have an 
extended discussion off-line. Additionally, here are two issues on which 
I'd love some comments:

1. The histograms of the raw pm & mm values are bimodal (see 
ftp://keck1.biotec.uiuc.edu/pub/Drnevich/Ecoli/ for output from 
signalDist). The most extreme array (# 9) had some problems with the 
labeling and hybridization, but it does appear to be at the end of a 
continuum. Looking through the Bioconductor archives, others have commented 
that bimodality may not be a problem in itself, but I wonder if anyone else 
has seen data this extreme or would be suspicious? The five arrays that 
different than the rest are from an different experimental treatment, which 
leads me to believe that this might be real data...

2. Applying bg.adjust.gcrma to the data results in almost zero expression 
for all pm probes! (see link above). I doubt that all the expression is due 
to non-specific binding and/or probe affinity. Is it likely that the 
different labeling of prokaryotic samples (N-terminus) would
render gcrma background correction unusable?

Thanks so much for any comments!
Jenny


Jenny Drnevich, Ph.D.

Functional Genomics Bioinformatics Specialist
W.M. Keck Center for Comparative and Functional Genomics
Roy J. Carver Biotechnology Center
University of Illinois, Urbana-Champaign

330 ERML
1201 W. Gregory Dr.
Urbana, IL 61801
USA

ph: 217-244-7355
fax: 217-265-5066
e-mail: drnevich at uiuc.edu