[BioC] Number or probes per target. Nimblegen arrays.

[Joern Toedling]

Hello,

my feeling and previous experience is that, with 60mers, you can get 
reasonable estimates for transcript concentration from a few probes 
only, as long as the 60mers are sufficiently specific for that 
transcript and distributed all over the array platform and not next to 
each other. In fact, Illumina uses one single 50mer per transcript for 
their beadarrays, but that one in about 30 copies all over the array to 
the effect that you get a very precise measurement for that oligo.
In your case, you probably want to be sure that the 60mers are specific 
for the genes/transcripts, have comparable GC contents, and no (partial) 
matches to other genes. See the paper by Rennie, Hoyle et al. in BMC 
Genomics for a recent, good paper on the effect of mismatches in long 
oligos.
The development version of Biostrings provides good, fast tools for 
assessing (partial) reporter matches in a genome. Nimblegen probably 
does a good job with the probe selection, but if you conduct such an 
assessment yourself beforehand, you will know what kind of 
cross-hybridization you could expect to see later in your data.

Regards,
Joern

J.delasHeras at ed.ac.uk wrote:
>
> Hi everyone,
>
> not especifically a BioC question this time... but I am hoping to use 
> the collective experience regarding a particular array design.
>
> We're considering using human expression arrays from Nimblegen. 
> Essentially, we're not sure which format to go for:
>
> 1) 6-20 60-mers per target
> 2) 2-7 60-mers per target
>
> The design #2 uses less probes per target, allowing to have 4 times a 
> 72k array on one slide (which can be hybridised with separate 
> samples), rather than 1 time a 385k array. So it's a smaller array but 
> being able to hyb 4 samples in one array reduces costs quite a bit... 
> so it's attractive, and my feeling is that it shouldn't be too 
> problematic to reduce teh number of probes per gene to this level... 
> but it's just a "feeling", having little experience on this type of 
> arrays.
>
> Any thoughts about this?
>
> Jose
>

-- 
Joern Toedling
EMBL - European Bioinformatics Institute
Wellcome Trust Genome Campus
Hinxton, Cambridge CB10 1SD
United Kingdom
Phone  +44(0)1223 492566
Email  toedling at ebi.ac.uk