Informative genes in NGS expression data

[Amative]

Hello all,

Say I have NGS gene expression data matrix (6 samples, ~ 30K genes). I am trying to perform some clustering. I was reading a paper talking about feature extraction techniques and how they help in extracting "Informative genes" before applying any clustering or classification algorithms. This paper was based on microarray expression data. Anyone here can guide to similar techniques that can be used to reduce the number of 30K genes to those informative ones.

Thanks a lot

[Richard Finney]

Consider normalization (i.e. using FPKM or such). It may already have been done for you.
Try filtering on some cutoff for variance. Keep most variant. You can also sub-set into groups or "ontologies", i.e. Kinome or chromodomain or whatever. Don't be surprised if you have different pictures (clusterings) for different cutoffs or subsets. One problem with clustering is that people can apply different filters until they the results they want.

[Amative]

Thanks Richard, I have already calculated the RPKM.
filtering on some cutoff for variance --> Any recommended tools, links ... ??
Regarding the missing values in the RPKM matrix, will the variance filtration take care of it or do is there anyway to deal with it?