Single-sample expression quantification

[n.o.worm]

Greetings--
Any suggestions for comparing expression levels of genes within a single sample? I want to look at relative expression of genes within a gene family in a single sample--is there a way to correct for length difference, or can I really only approximate using raw read counts? Any suggestions are greatly appreciated.

[mbblack]

In the simplest sense, you could just use RPKM values to compare. The problem is, for such a within sample comparison, technical variation becomes the issue, especially for low read abundance transcripts.

When comparing across biological replicates, biological variation typically swamps technical variation so one can dismiss technical variation as being (relatively) insignificant. In your case though, technical variation is a significant concern as your estimate of a particular transcript's abundance (again, especially for low abundance features) may change greatly across technical reps.

So to do what you wish with any confidence, you would need multiple technical replicates of that single sample, to have some actual measure of how well (or not) you have actually characterized the mean expression level of any given transcript or gene.