[BioC] No changes between Groups using GeneST Arrays from Affymetrix

[Mike Walter]

Hi Benjamin,

I don't have an explanation of your problem but some thoughts maybe. I'd agree with you that the bioinformatics is not the problem with your arrays. I'd rather think it is the sample processing part. 

First of all, maybe there is not much going on in your experiment (at least keep it as an option in mind). If you have some candidate transcripts which are highly likely to change, you can check them with qPCR. If they show regulation in the real-time PCR and not  on the array just repeat the experiment and discard the old hybridizations. We had some issues with the HuGene 1.0 ST arrays in the past. There were some batches of the WT Target Labelling Kit which gave strange results about a year ago. We found mainly snoRNAs regulated and nothing else. And the samples were treated with HDAC inhibitors, so you would expect a lot of transcriptional changes. After talking to the Affy FAS they admitted that there was a general problem with the kits and they replaced everything. When we repeated the experiment the results looked fine. How does the QC of your gene arrays look like if you compare them to gene arrays from your contact in Kiel or Berlin?

We also experienced some problems with the fluidics. In a larger project a large number of arrays showed bad QC (a huge shoulder in feature intensity density plots, higher overall intensities and high feature RLE). And all bad arrays were processed in the same fluidics station. We figured out that the problem always occured when the fluidics was used before to stain prokariotic arrays. Careful cleaning and two rounds of shutdown solved the problem. Sometimes the first batch after bleaching of the fluidics also showed spatial problem in the images (but not restricted to gene arrays). So I'd repeat a small experiment 3vs3 on gene arrays and 3'IVT arrays. If the results differ I'd talk to your Affy support, show them the results and get an replacement for the chips.

I hope this was of any help. I'd be really interested if you can nail your problem down.

Kind Regards,

Mike






------------------------


Hi guys,

 

although this is probably no problem concerning analysis maybe some of you
have an idea what might be going wrong here.

 

We are observing a strange problem with GeneST Arrays from Affymetrix.
Independent from the experiment design and number of arrays we hardly
observe any differentially expressed genes using the "gene level" analysis.
Maybe one to five genes if many and that would be without any multiple
testing correction. Even genes we know that they are regulated  or even
knocked out don't display any significant change. And we are talking about
cell culture experiments in some cases. Switching to "exon level" returns
some more candidates and our knowledge about the experiment tells us they
might be true positives but by the regulation strength and total number of
regulated candidates they might as well be false positives due to the much
higher number of probesets.

 

Now the issue worrying us is, that we performed several smaller studies with
the GeneST arrays know, mouse AND human, and didn't get a single study yet
with remarkable results. We never observed such problems with the "old"
human and murine whole genome arrays and all we get to hear from Affymetrix
is ".well other groups report the GeneSTs are running much better and more
sensitive than the old chips.and they are more robust .".

 

I don't think the statistical part is making the problems because we tried
normalization with RMA and PLIER (I didn't try VSN yet, maybe I should try
that) and I would expect a simple non-corrected Welch test and a screening
of potential targets by eye-balling to find at least a bunch of hits. But no
chance. And we are sticking to the standard kits and protocol from
Affymetrix in the lab. I already had a chat with the guys from Charite and
Max-Planck in Berlin and from the Affymetrix Core Facility in Kiel, Germany.
They are all doing just the same thing and getting nice results. We couldn't
find a step we are doing significantly different on computer or in lab.

 

So what I am interested in is: Does anybody of guys with experience with the
GeneST arrays have an idea what might be causing the problem on our side,
either already in lab or afterwards in the analysis? 

 

Best regards,

 

Benjamin

 

 

PS: I know the best step to go forward from here is performing the same
experiment on one of the old whole genome and on GeneST in parallel to check
the result. Still maybe somebody has an idea before we start with that.

 

 

___________________________________________

Benjamin Otto, PhD

University Medical Center Hamburg-Eppendorf

Institute For Clinical Chemistry

Campus Forschung N27

Martinistr. 52,

D-20246 Hamburg



Tel.: +49 40 7410 51908

Fax.: +49 40 7410 54971

___________________________________________





-- 
Pflichtangaben gemäß Gesetz über elektronische Handelsregister und Genossenschaftsregister sowie das Unternehmensregister (EHUG):

Universitätsklinikum Hamburg-Eppendorf
Körperschaft des öffentlichen Rechts
Gerichtsstand: Hamburg

Vorstandsmitglieder:
Prof. Dr. Jörg F. Debatin (Vorsitzender)
Dr. Alexander Kirstein
Joachim Prölß
Prof. Dr. Dr. Uwe Koch-Gromus

	[[alternative HTML version deleted]]
-------------- next part --------------
_______________________________________________
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