[BioC] Question for NGS seq data (time couse)

[Naomi Altman]

Dear Xing,
Last time I looked, edgeR, DESeq, etc were meant to test differences 
in means, which requires replication.

For a time course without replication, you probably need a 
regression-type analysis - e.g. fitting a negative-binomial 
regression on time for each gene.  Then the deviations from the model 
are used to assess the statistical significance of the pattern.  To 
do this, you need an a priori idea of the pattern - e.g. cosine waves 
for periodic genes and polynomials for linear or slightly curved genes.

An ad hoc method would be to do some expression clustering based on 
e.g. Spearman correlation to cluster the genes into 3 or 4 clusters 
with roughly similar shapes.  Then do regression on an appropriate 
basis in each cluster, using a generalized linear model.

--naomi

At 02:23 PM 3/2/2011, Xing Li wrote:
>Hi,
>
>I have a question regarding the NGS (RNA-Seq) analysis.
>
>We have a time course data regarding embryo development: 8 time points from
>different day mouse embryos and did RNA-seq on these samples. But we  have
>only one sample for each time point.
>
>I knew that there are some bioc packages for NGS analysis (edgeR, DESeq,
>etc). Is any package or method good for one-sample-per-time-point data?
>
>Thank you.
>
>Xing
>
>         [[alternative HTML version deleted]]
>
>_______________________________________________
>Bioconductor mailing list
>Bioconductor at r-project.org
>https://stat.ethz.ch/mailman/listinfo/bioconductor
>Search the archives: 
>http://news.gmane.org/gmane.science.biology.informatics.conductor