[BioC] comparison of two sets of developmental study

[Steve Lianoglou]

Hi,

On Mon, Apr 19, 2010 at 10:54 AM, Yan Zhou <Yan.Zhou at fccc.edu> wrote:
> Hi, Steve,
>
> Thank you for your reply. Sorry that I didn't make it clear. My interests
> are two layers:
>
> 1. Genes changing along development in tissue A  and B seperately; More
> precisely, genes change in A along development; Same as in tissure B;

It sounds like you want to call genes that are differentially
expressed along your time series.

You can use limma for that, but there are also several BioC packages
that specifically deal with timecourse data, such as:

* maSigPro: http://bioconductor.org/packages/2.5/bioc/html/maSigPro.html
and http://bioinformatics.oxfordjournals.org/cgi/content/full/22/9/1096

* betr: http://bioconductor.org/packages/2.5/bioc/html/betr.html

The vignettes that accompany them have references to publications you
can read to figure out what others are doing.

> 2. Anti-correlated expression patterns as you mentioned in the two tissues
> in order to find key players which drive the two development differently.

Assuming you have expr.a, and expr.b, which are the normalized
expression datasets (rows=genes, cols=times) for tissue a and b,
respectively, I reckon you can simply run cor.test along each row-pair
from the two matrices and look for highly negative correlation
coefficients w/ significant p.values as a start, no? For example:

R> cors <- lapply(1:nrow(expr.a), function(i) {
  cor.test(expr.a[i,], expr.b[i,])
})
R> interesting <- which(sapply(cors, function(x) x$p.value < 0.05 &&
x$estimate < 0))

`interesting` will have the indices of rows in the expression matrix
that have uncorrected p-values < 0.05 and are negative.

-steve

-- 
Steve Lianoglou
Graduate Student: Computational Systems Biology
 | Memorial Sloan-Kettering Cancer Center
 | Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/contact