[BioC] Poisson noise with respect to digital gene expression

[Dave Tang]

Dear list,

I have a very basic question regarding Poisson noise or shot noise with  
respect to differential expression analysis of digital gene expression  
data. As an example, if one source of RNA was made into 7 identical  
RNA-Seq libraries and then sequenced on 7 individual lanes, when comparing  
the sequence data we would still expect differences between the gene  
counts due to shot noise.

Using the rpois() function I generated some data that follows a Poisson  
distribution and checked the data distribution in terms of standard  
deviations. In contrast to the 68-95-99.7 rule for the normal  
distribution, I observed percentages of 73-96-99.6.

So in my digital gene expression example above, if I calculated the mean  
across the 7 technical replicates for one particular gene, I can expect  
that the counts for each respective library should lie within one or two  
standard deviations of the mean. Is this what the Poisson or shot noise is  
referring to?

Many thanks,


-- 
Dave