[Arrays] microarray data analysis

Igor Chernukhin via arrays%40net.bio.net (by igorc from essex.ac.uk)
Tue Feb 7 06:00:45 EST 2012


Hi Harish -
Thanks for your reply.

On Tue, 2012-02-07 at 10:25 +0530, Harish Rotti wrote:
> Hi Igor,
> 
> 
> I assume the array what you have used is a Nimblegen Expression array
No, this is a *tilling* array.
( http://en.wikipedia.org/wiki/Tiling_array )

>   and you have performed two colour hybridization with a dye swap to
> address your question.. In order to estimate up-regulated genes in a
> given condition(condition is nothing but logCy5-logCy3) analyse the
> data with your dye swap (logCy3-logCy5) data in Limma. Theoretically
> the up regulated probes in the usual experiment should be down
> regulated in dye swap,however due to dye effect it will not be . Thats
> why limma and other software like Gene Spring performes Dye Swap
> Normalization and gives Upregulated and down regulated gene list. More
> details is present in the  free access paper entitled as  "Assessing
> the efficiency of dye-swap normalization to remove systematic bias
> from two-color microarray data". To present the data you can use
> Volcano Plot.

There is no problem with two-color stats. We use limma
normalizeWithinArrays() and normalizeBetweenArrays() (or vsn2()),
alternatively commercial Nimblegen RMA will do it all, there is no
problem with this. The problem that we have output in scaffold
coordinates. For example limma eBayes() output will look like this:

------------------------------------------------------------------------
        GENE_EXPR_OPTION                 PROBE_ID POSITION    X    Y
Status
504944            BLOCK1  SCAFFOLD_154FS000176133   176133  247  179
Probe
1550318           BLOCK1    SCAFFOLD_4FS002269602  2269602  354 2862
Probe
1930401           BLOCK1  SCAFFOLD_791FS000015470    15470  238 1138
Probe
1424296           BLOCK1  SCAFFOLD_426FS000011971    11971 1043  335
Probe
                              ID     logFC  AveExpr         t
P.Value
504944   SCAFFOLD_154FS000176133 -4.350414 12.82226 -64.53841
2.787723e-12
1550318    SCAFFOLD_4FS002269602 -5.005838 12.47949 -62.02124
3.846196e-12
1930401  SCAFFOLD_791FS000015470 -4.319709 12.11767 -61.93263
3.890941e-12
1424296  SCAFFOLD_426FS000011971 -4.310220 11.14722 -61.86019
3.927954e-12
           adj.P.Val        B
504944  1.051002e-06 17.02628
1550318 1.051002e-06 16.86694
1930401 1.051002e-06 16.86109
1424296 1.051002e-06 16.85628
-------------------------------------------------------------------------

We have stats for each probes (~50bp each) mapped into scaffolds (this
is genome-wide, it contains both coding and non-coding sequences). What
people want to know if we can (somehow) estimate a *mean* level of
expression per transcripts (or exons) using this data. Obviously there
will be multiple probes per single exonic region and those probes will
vary in expr and significance. The question is can we have a mean of
logFC values (or AveExpr) of probes laying into this region presented as
this value? 


> May I know how you calculated fold difference? As for as my
> understanding, fold difference is calculated in the following
> condition.....
diffmean.stat from R-st-package
(Breitling, R., et al. 2004. Rank products: a simple, yet powerful, new
method to detect differentially regulated genes in replicated microarray
experiments.  FEBS Letters *573*:83-9.)

> Suppose you are trying to see efficacy of drug,  then before treatment
> isolate mRNA and label with cy3 as well as cy5 and after treatment
> label mRNA with cy5. Do two hybridization one Cy3 of Untreated vs Cy5
> of untreated, second Cy3 of Untreated vs Cy5 of Treated. So in this
> you have Cy3 Chanel is being same for both expt and you can estimate
> Fold change difference. But in your case, expt is performed some what
> - Cy3 untreated  vs Cy5 treated with dye swap, where measuring fold
> change difference is not logically correct.
> 
> 
> Thanks  
> 
> 
> Thank you
> 
> 
> 
> 
>  
> with regards
> Harish Rotti
> Ph.D Scholar
> Department of biotechnology
> Manipal life sciences centre
> Manipal
> 
> 
> ______________________________________________________________________
> From: Igor Chernukhin <igorc from essex.ac.uk>
> To: arrays from magpie.bio.indiana.edu 
> Sent: Monday, 6 February 2012 3:12 PM
> Subject: [Arrays] microarray data analysis
> 
> 
> 
> Hi All -
> I would greatly appreciate if you give me advice on microarray data
> interpretation. 
> We have a custom tilling microarray from Nimblegen for a newly
> sequenced
> genome of (some) organism. The genome is poorly annotated yet and all
> probe annotations are in scaffold coordinates. It is dye-swap array
> hybridized with mRNA from two different biological conditions. We have
> processed the data with nimblegen commercial RMA software and also
> have
> stats made with limma/vsn2 for each probe: fold difference (pmean),
> fdr.
> The research group who studies this organism wants rather simple
> answer
> like which transcripts are upregulated at which conditions and present
> it in a ‘simple’ format. RMA output is not really helpful because it
> gives all in scaffold coordinates. However we can see the difference
> in
> probe values corresponding to transcript coordinates (annotated from
> currently available gff) and of course it varies within the region
> from
> significant to nonsignificant. Would it be appropriate to calculate
> the
> mean of pmean values for each region corresponding to transcript (or
> exons) and give it as the answer? Is there any other way of doing it…?
> 
> Many thanks
> -Igor
> 
> _______________________________________________
> Arrays mailing list
> Arrays from net.bio.net
> http://www.bio.net/biomail/listinfo/arrays
> 
> 
> 
regards
-Igor



More information about the Arrays mailing list