[BioC] Interpreting results...

[James Wettenhall]

On Thu, 23 Oct 2003, Marcelo Luiz de Laia wrote:

> Hi everyone.
> 
> I have the following layout:
> 
> SlideNamber	FileNameCy3	FileNameCy5	Cy3	Cy5
> 1	lamina1_1_selected.txt	lamina1_0_selected.txt	BCYE	XDM
> 2	lamina2_1_selected.txt	lamina2_0_selected.txt	XDM	BCYE
> 3	lamina3_1_selected.txt	lamina3_0_selected.txt	XDM	BCYE
> 
> And I have these results (fictitious):
> 
> .Field	Meta.Row	Meta.Column	Row
> 4432	XDM/BCYE	3	2	25
> 2955	XDM/BCYE	2	2	28
> 5203	XDM/BCYE	4	1	25
> 4510	XDM/BCYE	3	2	28
> 1533	XDM/BCYE	1	2	32
> 4562	XDM/BCYE	3	2	31
> 5137	XDM/BCYE	4	1	23
> 
> Column	Gene.ID	M	t	P.Value	B
> 16	SemP01	-3.00	-61.01	0.00	7.96
> 3	ty0012	-2.83	-51.54	0.00	7.78
> 19	SemP08	-2.73	-51.46	0.00	7.71
> 22	SemP12	-3.55	-51.35	0.00	7.62
> 21	ty2196	-3.97	-41.90	0.00	5.68
> 2	SemP38	-2.60	-41.70	0.00	5.48
> 1	ty0559	1.66	41.09	0.00	5.03
> 
> Which is the interpretation for the gene 4432? Inside of this layout it was
> up or down regulated?
> 
> And the gene 5237? Was it up or down?
> 
> Thanks very much
> 
> Marcelo

Marcelo,

Given that you have a toptable, I assume that at some stage you 
were successful in specifying a design matrix.  A quick look at 
your data suggests design=c(1,-1,-1), meaning that you are 
estimating one parameter (for each gene) which is 
log2(XDM/BCYE).

A gene such as 4432 which has a negative M value in the toptable 
and negative moderated t statistic is down-regulated in XDM 
compared with BCYE with a log2 fold change given by the M value.  
A (moderated) t statistic which is large in magnitude means a 
high confidence of differential expression because there is good 
consistency between replicates as well as a significant log2 
fold change for that gene.

5137 is upregulated in XDB compared with BCYE with high 
confidence (for similar reasons).

The B statistic estimates the log odds of differential 
expression of a gene given its M value.  So if it is greater 
than zero, then you have more than a 50/50 chance that the gene 
is differentially expressed.

Hope This Helps,
James