[BioC] DESeq multifactor GLM Error attributed to zero counts (coded as NA) in dataset?

[Simon Anders]

Hi Steve,

On 08/10/2011 10:35 PM, Steve Vollmer wrote:
> I get the following error message in the new (beta) version of DESeq
> when I try to implement the fitNbinomGLMs function
> [e.g.fitNbinomGLMs(cds, count ~ treatment + condition)].
>
> Error in glm.fit(x = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0,  :
>    NA/NaN/Inf in foreign function call (arg 1)
>
>
> This error message occurs regardless of which method (pooled, blind, etc.) that I use to estimateDispersions.

Occasionally, glm.fit fails to converge and then the function throws an 
error. Just yesterday, I've updated DESeq and wrapped the call to 
glm.fit in a 'try' statement, so that the function finishes and you just 
get a few NAs as p values from these cases.

Let me know if the problem persists in DESeq version 1.5.24.

   Simon