[BioC] bgx on data without replication

[Ernest Turro]

On 10 Jun 2010, at 17:54, Ina Hoeschele wrote:

> Hi, I have been analyzing an affy microarray dataset with 2 treatments and 1 control, which do not have replication (so total of 3 chips), with several methods that do not require replication, including bgx. BGX seems to give the poorest results which I did not expect, so I am wondering whether I am doing something wrong. I ran BGX as follows:
>> library(bgx)
>> Data <- ReadAffy()
>> bgx(Data,burnin=9216,iter=20480,output="minimal")
>> bgxOutput <- readOutput.bgx("run.1")
>> rankedGeneList1 <- rankByDE(bgxOutput,conditions=c(1,2),absolute=TRUE)  #comparing control to treatment 1
>> rankedGeneList2 <- rankByDE(bgxOutput,conditions=c(1,3),absolute=TRUE)  #comparing control to treatment 2
>> plotDEHistogram(bgxOutput, conditions=c(1,2),normalize="none")  #I am doing this to get the number of differentially expressed probesets
> Number of differentially expressed  genes (left):  1 
> Number of differentially expressed  genes (right):  590 
> Total number of differentially expressed genes:  591 
>> plotDEHistogram(bgxOutput, conditions=c(1,3),normalize="none")
> Number of differentially expressed  genes (left):  31 
> Number of differentially expressed  genes (right):  106 
> Total number of differentially expressed genes:  137 
> 
> This then gives me the lists of differentially expressed genes:
>> list1 <- rankedGeneList1[1:591,]
>> list2 <- rankedGeneList2[1:137,]
> 

Could you try using normalize="loess" ? The histogram should be centred on 0.5 and have peaks near 0 and 1. The spline fit should be smooth. What do your plotDEHistogram() plots look like?

Thanks
E



> Thanks for any comments ...
> Ina