[BioC] CGH analysis without genome positions

[Sean Davis]

On Tue, Apr 6, 2010 at 6:53 PM, adam_pgsql <adam_pgsql at witneyweb.org> wrote:
>
> On 6 Apr 2010, at 19:15, Sean Davis wrote:
>
>> On Tue, Apr 6, 2010 at 1:58 PM, Sean Davis <seandavi at gmail.com> wrote:
>>> On Tue, Apr 6, 2010 at 1:47 PM, adam_pgsql <adam_pgsql at witneyweb.org> wrote:
>>>>
>>>> Hi,
>>>>
>>>> I am trying to do some CGH analysis with Agilent arrays, but all the analyses methods seem to require genome position information. Does anyone know of any packages that will call genes as present/absent without the genome position?
>>>>
>>>
>>> I don't think of CGH analysis as "present/absent", but perhaps I am
>>> not clear on what you mean by CGH analysis.  For Agilent arrays,
>>> presumably you have two colors, one representing the sample and the
>>> other the reference.  Simply make a ratio and then rank the probes
>>> based on that.
>>
>> I'm making an assumption here that you are using some custom array
>> based on an organism with no assembled genome.  If there is an
>> assembled genome, then you should map your probes to the genome using
>> an alignment tool (blast, blat, etc.) and use those alignments for
>> more standard CGH analysis.
>
> Thanks Sean for your reply.
>
> This is a custom bacterial pan-genome array. The problem is that many of the oligos target genes found in unfinished genome sequences (not the reference strain) and as such I don't really have a genome position. Also due to the nature of bacterial genomes when i hybridise DNA from unsequenced strains there is no guarantee that the gene arrangement would be exactly the same as the sequenced reference strain.
>
> in terms of "present/absent" i would like to score each gene sequence represented on the array as present or absent in the test strain. I guess this could be done by ranking the ratios and determining some cutoff for presence or absence, but the question is are there any tools that provide a more statistically sound approach to suggesting a good cutioff value to use?
>

Hi, Adam.

There are many ways to go here, but one would really need to know the
experimental design in more detail.  If you have replicates, then
there are MANY statistical methodologies that could be applied to find
differences between the reference and the test.  Any gene expression
hypothesis testing packages could probably be applied.

Sean