[BioC] Deciding on a cut off after QC

[Gordon Smyth]

At 01:48 PM 17/05/2005, Ankit Pal wrote:
>Dear Adai,
>Thank you for the detailed response.
>Correct me if I'm wrong.
>What you are saying is that after applying the "fdr"
>I give a cut off of 0.05 for the p-value and write all
>those spots into an object (positives).
>That means, I do  not consider the B values.

This is one way to proceed. Have you read the section in the User's Guide 
on "Statistics for differential expression" which discusses which statistic 
to use?

>In essence, it is just a one sample t test afetr
>adjusting the p-value and taking into consideration
>all those spots that have a p-value < 0.05?
>
> From my code in a previous mail to Gordon Smyth, I
>have done a quality control (QC) using parameters and
>threshold values prescribed by genepix.
>I know from experience, a large number of spots do not
>get through the filter.
>As I understand from LIMMA, a weight of "0" is given
>to any spot that does not get through the QC filter.
>How do I exclude these spots from the final result
>summary file?

These spots are already excluded. You seem to be confused by the fact that 
excluding a spot does not exclude the corresponding probe, since there are 
four spots for each probe. You will get a valid t-statistic and p-value for 
a probe if any one of the spots for that probe get a positive weight. (Note 
you can get a Bayesian t-statistic even for a probe with no residual df.)

>I need to get a set of significantly differentially
>expressed genes that have got through the QC filter.
>How do I do it using LIMMA?

This is what you have been getting all along.

Gordon

>Thank you,
>-Ankit