[BioC] time course experiment

[Gordon K Smyth]

That's right.

Gordon

On Wed, 23 Jul 2014, Rao,Xiayu wrote:

> Hi, Gordon
>
> Thank you very much for information. I see what you meant now. I also checked to see the difference between topTable and topTableF and became more familiar with the output.
> (1) topTableF has moderated F-statistics, whereas topTable uses moderated t-statitics.
> (2) The log2 fold changes for a specific comparison are the same (i.e. both are -1.94 for wt.6h-wt.0h), and the average expression values are the same for a probeset.
> (3) Although for the 1st test only 2 contrasts were specified, the topTableF(fit2,number=Inf,p.value=0.05) gene list should include all the genes that are significantly different either between 0h and 6h, or between 0h and 24h, or between 6h and 24h in the wt. That's why it is said in the user guide that any two contrasts between the three times would give the same result. Am I correct?
>
> ###1. Test for gene change over time in wt
>> cont.129 <- makeContrasts("wt.6h-wt.0h"," wt.24h-wt.6h",levels=design)
>> fit2 <- contrasts.fit(fit, cont.129)
>> fit2 <- eBayes(fit2)
>> topTableF(fit2)
> PROBEID	SYMBOL	GENENAME	                                               ENTREZID	wt.6h-wt.0h	wt.24h-wt.6h	AveExpr	F	P.Value	adj.P.Val
> 1436717_x_at	Hbb-y	hemoglobin Y, beta-like embryonic chain	15135	-1.94629866	-6.329636145	11.0091773	2601.286661	3.95E-16	6.07E-12
>
> ###2. Test for gene change between 2 time points
>> cont.all <- makeContrasts("mu.0h-wt.0h"," mu.6h-wt.6h"," mu.24h-wt.24h"," wt.6h-wt.0h"," wt.24h-wt.6h","mu.6h-mu.0h"," mu.24h-mu.6h",levels=design)
>> fit2 <- contrasts.fit(fit, cont.all)
>> fit2 <- eBayes(fit2)
>> topTable(fit2,coef=4)
> PROBEID	SYMBOL	GENENAME	                                               ENTREZID	logFC	AveExpr	t	P.Value	adj.P.Val	B
> 1436717_x_at	Hbb-y	hemoglobin Y, beta-like embryonic chain	15135	-1.94629866	11.0091773	-16.22147355	2.45E-09	8.47E-07	12.0412676
>
>> topTable(fit2,coef=5)
> PROBEID	SYMBOL	GENENAME	                                               ENTREZID	logFC	AveExpr	t	P.Value	adj.P.Val	B
> 1436717_x_at	Hbb-y	hemoglobin Y, beta-like embryonic chain	15135	-6.329636145	11.0091773	-52.75450649	3.39E-15	5.14E-11	20.36773133
>
>
> Thank you for your sharing! I am willing to learn.
>
> Xiayu
>
>
>
>
>
>
>
> -----Original Message-----
> From: Gordon K Smyth [mailto:smyth at wehi.EDU.AU]
> Sent: Tuesday, July 22, 2014 6:04 PM
> To: Rao,Xiayu
> Cc: Bioconductor mailing list
> Subject: RE: time course experiment
>
> On Tue, 22 Jul 2014, Rao,Xiayu wrote:
>
>> Hi, Gordon
>>
>> Thank you for letting me know. As Jim suggested, I would just include
>> everything in one command as below to make it simple.
>
> It's just convenience.  It can be done either way.
>
>> I now understand that the difference between testing for changed gene
>> over time (the trend) and testing for diff genes between two exact
>> time points is
>
> Actually there is no difference between the two from limma's point of view.  You simply compute any contrast of interest to you and then test for DE for that contrast.
>
>> (1) to make different contrasts and
>
>> (2) to use topTableF(fit) to extract the gene list for the former and
>> to use topTable(fit, coef=1 or any other number) for the later.
>> (Correct me if I am wrong, thank you)
>
> Actually, given the way you have computed your contrasts, you probably want to specify coef for any topTable.
>
> Gordon
>
>> contrast <- makeContrasts("mu.0hr-wt.0hr", "mu.6hr-wt.6hr",
>> "mu.24hr-wt.24hr","wt.6hr-wt.0hr", "wt.24hr-wt.6hr",
>> "wt.24hr-wt.0hr","mu.6hr-mu.0hr", "mu.24hr-mu.6hr",
>> "mu.24hr-mu.0hr",levels=design)
>>
>> Thanks,
>> Xiayu
>>
>>
>>
>> -----Original Message-----
>> From: Gordon K Smyth [mailto:smyth at wehi.EDU.AU]
>> Sent: Tuesday, July 22, 2014 5:26 PM
>> To: Rao,Xiayu
>> Cc: Bioconductor mailing list
>> Subject: time course experiment
>>
>> Dear Xiayu,
>>
>> I don't quite see the problem.  It all seems straightforward.  All the
>> contrasts you have proposed seem simple and sensible.  There is no
>> need nor possible advantage in subsetting the data.
>>
>> Best wishes
>> Gordon
>>
>> PS. I haven't included you original post in my reply because there
>> were so many non-standard characters imbedded in it.

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}