[BioC] All equal high p-values from limma topTable . What to do ? Need help...

[James W. MacDonald]

Giulio Di Giovanni wrote:
> Hi I'm looking at the results of an analysis on 4 S. Cerevisiae cDNA arrays, 
> (2 of these have a dye-swap).
> I followed the example in limma user's guide (that if I'm not wrong is 
> exactly my case).
> 
> I obtain a topTable of genes of this type
> 
> 	Block	Row	Column	ID	Name	M	A	t	P.Value	B
> 4030	30	11	11	YLR162W	:::::1:	1.624122241	6.560234285	4.467743568	0.997257765	-2.0502737
> 1254	10	4	11	YDR166C	SEC5:::::1:	-0.56584704	5.143049435	-3.824556839	0.997257765	-2.61592631
> 1113	9	4	1	YDR409W	:::::1:	1.016356426	4.282139218	3.988554403	0.997257765	-2.683464218
> 2115	16	10	5	YKL209C	STE6:::::1:	-0.540528098	7.340027942	-3.533312325	0.997257765	-2.686586054
> 5022	37	16	11	YPR070W	:::::1:	-0.493446684	6.765075537	-3.415141086	0.997257765	-2.77744221
> 3642	27	14	3	YOL045W	:::::1:	-0.678996258	5.185557216	-3.448815561	0.997257765	-2.865258763
> 1602	12	14	3	YNL253W	:::::1:	-0.542907069	6.019893924	-3.427152231	0.997257765	-2.880360788
> 4177	31	13	1	YNL221C	POP1:::::1:	-0.49777162	6.787712846	-3.274291885	0.997257765	-2.8886384
> 
> 
> Where all the p-values are 0.997257765. I read in the topTable help  that 
> "if there is no good evidence for differential
>      expression in the experiment, that it is quite possible for all
>      the adjusted p-values to be large, even for all of them to be
>      equal to one."
> 
> I'm quite astonished ... and now ?
> This fact implies that is not a good experiment ? Or that data were not well 
> preprocessed ? Or maybe that for that experiment there are no genes 
> significantly differently expressed ?

It means pretty much what the help file says, that there is no good 
evidence for differential expression for any of your genes. Why that 
might be is a different question with many possibilities. These include

1.) Very small differences - looking at your topTable, there are only 
two genes with > 2-fold difference. Maybe there really aren't any 
differences between the samples.

2.) Noisy data - if the variability between the chips is high, then you 
will need larger differences in order to gain statistical significance. 
Alternatively, you may just need more data, which will increase your 
power to detect differences.

3.) Incorrect normalization - if the normalization was not done 
correctly, you may not be accounting for some non-biological 
variability, which will result in increased variability.

4.) Coding mistakes - maybe you aren't doing the comparisons you think 
you are doing.

I don't think you will be able to get much help here, because what you 
really need is for an experienced person to look at your data and code. 
This is not amenable to a listserv, so your best bet is to find a local 
statistician who might be able to help you out.

Best,

Jim


> 
> I'm analyzing that data from a Biomolecular lab, and I don't know what to do 
> and how to explain this...
> 
> I'll be very happy for any help or suggestion ... !!!!!
> 
> Thanks in advance,
> 
> Giulio...
> This
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor


-- 
James W. MacDonald, M.S.
Biostatistician
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109
734-647-5623


**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues.