Dear Alper
this sort of clustering is an explorative data analysis technique, as
such there is no single 'right' answer (although there are 'wrong' ones :)
So, what you suggest below is valid, although it could be suboptimal
since the clustering could be dominated by few very large, but unprecise
per-gene ratios especially when one of (Exp1-rpmk) or (Exp2-rpmk) is
small. Section 4.6 and Fig.5 of this paper suggest a remedy:
http://precedings.nature.com/documents/4282/version/2
Best wishes
Wolfgang
On 02/05/10 17:28, Alper Yilmaz wrote:
Hi,
There's very nice tutorial about using Bioconductor to calculate
clusters (of genes) with different algoritms using microarray data.
I would like to use the similar approach to calculate gene clusters
using RNA-Seq data. My question is, would it be okay to take the ratio
of RPMK values of Exp1 and Exp2 and use the natural log of that ratio
as if it's microarray data?
Let's say Exp1 is control and Exp2 is treatment. I have RPMK value for
each gene for both samples. Then, ratio of Exp2(RPMK) over Exp1(RPMK)
is calculated and then natural log of that ratio is calculated. And
from here on, use the tutorial to calculate gene clusters.
If log( Exp1-rpmk / Exp2-rpmk) is not a valid approach, what should I
do instead, so that I can use invaluable bioconductor tools that are
designed for microarray data mining, for RNA-Seq analysis?
Regards,
Alper Yilmaz
Post-doctoral Researcher
Plant Biotechnology Center
The Ohio State University
1060 Carmack Rd
Columbus, OH 43210
(614)688-4954
_______________________________________________
Bioc-sig-sequencing mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing
--
Wolfgang Huber
EMBL
http://www.embl.de/research/units/genome_biology/huber
_______________________________________________
Bioc-sig-sequencing mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing
