[BioC] duplicate correlation of 4 "within array replicates" limmaGUI

[Gordon K Smyth]

Dear Christine,

duplicate correlation in limma is only for replicate spots which are 
regularly spaced.  The spacing value is the number of spots between 
replicates.  Your value of 21 cannot be correct on an array with 
16 columns and 11 rows, and that is what limmaGUI is telling you.

You haven't actually told us how the 4 replicates for each mir are 
arranged on the arrays.

If your replicates really are regularly spaced, then limma commands 
readGAL() and getLayout(guess=TRUE) will probably tell you what the 
spacing is.

Best wishes
Gordon

> Date: Wed, 8 Oct 2008 20:07:09 +0200
> From: "Christine Voellenkle" <christine.voellenkle at gmail.com>
> Subject: [BioC] duplicate correlation of 4 "within array replicates"
> 	limma	gui
> To: bioconductor at stat.math.ethz.ch
> Message-ID:
> 	<b184b04b0810081107o1c98ab67hee91b203524835c8 at mail.gmail.com>
> Content-Type: text/plain
>
> Dear all,
>
> I perform 2 color hybridization, using microRNA and Exiqon slides.
> slide design: 4 replicates of 1 array
>                   each array consits of 8 subarrays,
>                   each subarray of 16 columns and 11 rows
>                   each mir has 4 replicates
>
> For analysis of the data I use limma GUI, to obtain the p-values of
> differentially expressed mirs I carried out loess normalization
> (backgroundsubstraction normexp, offset=10) and subsequently applied the
> Linear model fit, which works, as long as I treat  each replicate like a
> single mir.
> I would like to now also about the correlation of the replicates, so I
> entered as replicate number 4 and as distance 21 (according to
> "Bioinformatics+ computional statistics using R and Bioconductor" the
> distance corresponds to the number of spots that lies betweeen the
> replicates).
> But I get the following message:
>
> "Error in dim(M)<- c(spacing,ndups, ngroups, nslides)
> dims[product 5628] do not match the length of object [5632]"
>
> I rechecked the gal file and recounted on the image, 21 is the correct
> number (considering the column-distance, using the number of spots of row
> distance results in dims[product 0]).
> I read the limma guide and checked also the examples on the limma gui
> homepage, but it's always either no replicate or only 2 replicates next to
> each other.
> Do you know a solution?
>
> Also another question:
> IF one day I manage to do the duplicate correlation and given that I would
> use weights to identfy features with the flag "bad", how will the software
> handle this if one or more replicates are flagged bad? will it exclude the
> flagged bad replicate from the correlation?
> If yes, is there a way to put a minimum limit of 2 unflagged spots per mir,
> so that in case of only 1 unflagged replicate of a mir to exclude this mir
> from analysis?
>
> Thank you  a lot in advance!
> Greetings, Christine
>
>
>
>
>
>
>
>
> Dr. Christine Völlenkle, Ph.D.
> Research Laboratories-Molecular Cardiology
> I.R.C.C.S. Policlinico San Donato
> Via R. Morandi, 30
> 20097  S. Donato M.se (MI) Italy
> Phone: +39 02 52774 683 (lab)
>          +39 02 52774 533 (office)
> Fax:    +39 02 52774 666
> email: christine.voellenkle at gmail.com